WO2005026907A9 - Method and apparatus for computerized image background analysis - Google Patents

Method and apparatus for computerized image background analysis

Info

Publication number
WO2005026907A9
WO2005026907A9 PCT/US2004/029417 US2004029417W WO2005026907A9 WO 2005026907 A9 WO2005026907 A9 WO 2005026907A9 US 2004029417 W US2004029417 W US 2004029417W WO 2005026907 A9 WO2005026907 A9 WO 2005026907A9
Authority
WO
WIPO (PCT)
Prior art keywords
image
video
image region
background
video frame
Prior art date
Application number
PCT/US2004/029417
Other languages
French (fr)
Other versions
WO2005026907A2 (en
WO2005026907A3 (en
Inventor
Christopher J Buehler
Original Assignee
Intellivid Corp
Christopher J Buehler
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Intellivid Corp, Christopher J Buehler filed Critical Intellivid Corp
Priority to DE602004028136T priority Critical patent/DE602004028136D1/en
Priority to AU2004272178A priority patent/AU2004272178B2/en
Priority to AT04788655T priority patent/ATE474293T1/en
Priority to CA2538294A priority patent/CA2538294C/en
Priority to EP04788655A priority patent/EP1665127B1/en
Publication of WO2005026907A2 publication Critical patent/WO2005026907A2/en
Publication of WO2005026907A3 publication Critical patent/WO2005026907A3/en
Publication of WO2005026907A9 publication Critical patent/WO2005026907A9/en

Links

Classifications

    • GPHYSICS
    • G08SIGNALLING
    • G08BSIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B13/00Burglar, theft or intruder alarms
    • G08B13/18Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength
    • G08B13/189Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems
    • G08B13/194Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems
    • G08B13/196Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems using television cameras
    • G08B13/19602Image analysis to detect motion of the intruder, e.g. by frame subtraction
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/20Analysis of motion
    • G06T7/254Analysis of motion involving subtraction of images
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00Scenes; Scene-specific elements
    • G06V20/50Context or environment of the image
    • G06V20/52Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • GPHYSICS
    • G08SIGNALLING
    • G08BSIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B13/00Burglar, theft or intruder alarms
    • G08B13/18Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength
    • G08B13/189Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems
    • G08B13/194Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems
    • G08B13/196Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems using television cameras
    • G08B13/19602Image analysis to detect motion of the intruder, e.g. by frame subtraction
    • G08B13/19604Image analysis to detect motion of the intruder, e.g. by frame subtraction involving reference image or background adaptation with time to compensate for changing conditions, e.g. reference image update on detection of light level change
    • GPHYSICS
    • G08SIGNALLING
    • G08BSIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B13/00Burglar, theft or intruder alarms
    • G08B13/18Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength
    • G08B13/189Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems
    • G08B13/194Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems
    • G08B13/196Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems using television cameras
    • G08B13/19602Image analysis to detect motion of the intruder, e.g. by frame subtraction
    • G08B13/19608Tracking movement of a target, e.g. by detecting an object predefined as a target, using target direction and or velocity to predict its new position
    • GPHYSICS
    • G08SIGNALLING
    • G08BSIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B13/00Burglar, theft or intruder alarms
    • G08B13/18Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength
    • G08B13/189Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems
    • G08B13/194Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems
    • G08B13/196Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems using television cameras
    • G08B13/19639Details of the system layout
    • G08B13/19641Multiple cameras having overlapping views on a single scene
    • GPHYSICS
    • G08SIGNALLING
    • G08BSIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B13/00Burglar, theft or intruder alarms
    • G08B13/18Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength
    • G08B13/189Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems
    • G08B13/194Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems
    • G08B13/196Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems using television cameras
    • G08B13/19639Details of the system layout
    • G08B13/19652Systems using zones in a single scene defined for different treatment, e.g. outer zone gives pre-alarm, inner zone gives alarm
    • GPHYSICS
    • G08SIGNALLING
    • G08BSIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B13/00Burglar, theft or intruder alarms
    • G08B13/18Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength
    • G08B13/189Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems
    • G08B13/194Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems
    • G08B13/196Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems using television cameras
    • G08B13/19654Details concerning communication with a camera
    • G08B13/19656Network used to communicate with a camera, e.g. WAN, LAN, Internet
    • GPHYSICS
    • G08SIGNALLING
    • G08BSIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B13/00Burglar, theft or intruder alarms
    • G08B13/22Electrical actuation
    • G08B13/24Electrical actuation by interference with electromagnetic field distribution
    • G08B13/2402Electronic Article Surveillance [EAS], i.e. systems using tags for detecting removal of a tagged item from a secure area, e.g. tags for detecting shoplifting
    • G08B13/2451Specific applications combined with EAS
    • G08B13/246Check out systems combined with EAS, e.g. price information stored on EAS tag
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N5/00Details of television systems
    • H04N5/222Studio circuitry; Studio devices; Studio equipment
    • H04N5/262Studio circuits, e.g. for mixing, switching-over, change of character of image, other special effects ; Cameras specially adapted for the electronic generation of special effects
    • H04N5/272Means for inserting a foreground image in a background image, i.e. inlay, outlay
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N7/00Television systems
    • H04N7/18Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
    • H04N7/181Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast for receiving images from a plurality of remote sources

Definitions

  • the present invention generally relates to video surveillance, and more specifically to a computer-aided surveillance system for analyzing image backgrounds.
  • a simple closed-circuit television system uses a single camera connected to a display device. More complex systems can have multiple cameras and/or multiple displays.
  • One known type of system is the security display in a retail store, which switches periodically between different cameras to provide different views of the store. Higher security installations, such as prisons and military installations, use a bank of video displays each displaying the output of an associated camera. A guard or human attendant constantly watches the various screens looking for suspicious activity.
  • Computer-aided surveillance has been developed for various applications. For example, the military has used computer-aided image processing to provide automated targeting and other assistance to fighter pilots and other personnel. In addition, computer-aided surveillance has been applied to monitor activity in swimming pools.
  • a CAS system automatically monitors objects (e.g., people, inventory, etc.) as they appear in series of surveillance video frames.
  • objects e.g., people, inventory, etc.
  • One particularly useful monitoring task is tracking the movements of objects in a monitored area.
  • Methods for tracking objects, such as people, moving through an image are known in the art.
  • the CAS system can utilize knowledge about the basic elements of the images depicted in the series of surveillance video frames.
  • a video surveillance frame depicts an image of a scene in which people and things move and interact.
  • a video frame is composed of a plurality of pixels, often arranged in a grid-like fashion. The number of pixels in an image depends on several factors including the resolution of the camera generating the image, the display on which the image is presented, the capacity of the storage device on which the images are stored, etc. Analysis of a video frame can be conducted either at the pixel level or at the (pixel) group level depending on the processing capability and the desired level of precision.
  • Image regions can be categorized as depicting part of the background of the scene or as depicting a foreground object. In general, the background remains relatively static in each frame. However, objects are depicted in different image regions in different frames.
  • object extraction Several methods for separating objects in a video frame from the background of the frame, referred to as object extraction, are known in the art. A common approach is to use a technique called “background subtraction.” Of course, other techniques can be used. The locations of the objects are typically recorded in a list that is associated with the video frame.
  • a tracking method determines a correspondence between extracted objects in the current frame and extracted objects in the previous frame. This correspondence can be determined, for example, by using a predictive tracking method.
  • the CAS system predicts the location of an object in the current frame based on the known locations of the object in previous frames. Subsequently, the predicted object location is compared to the actual object location to establish correspondence.
  • Such a prediction is typically based on an algorithm that predicts likely object movement. For example, it can be assumed that objects move with constant velocity. More sophisticated techniques can, for example, verify that the colors of the objects match before determining a correspondence.
  • Examples of background features can include: (i) frequently being immediately “below” or at the bottom of objects in video frames, (ii) frequently having objects first appear in a series of video frames in the image region, (iii) having objects frequently leave a series of video frames from the image region, (iv) being capable of occluding other objects, etc.
  • Image regions having one or more of these features can be classified as being a floor, a portal, or an obstruction.
  • a human operator can assign these classifications to image regions based on these or other features.
  • a user interface such as that used in a digital paint program
  • an operator can "paint" classifications onto a video image in areas that correspond to floors, obstructions, and portals.
  • These classifications may not be mutually exclusive, so the different classifications can be represented with different "layers.” While painting these pixels might be trivial, it is a time consuming and error-prone task.
  • the human effort involved would be prohibitive, which would make a CAS system that uses this information difficult to deploy in large-scale installations.
  • the invention relates to a computer method of video analysis.
  • the method includes receiving image data for a plurality of video frames that depict a scene.
  • the scene exhibits at least one of a plurality of background features and each video frame is composed of a plurality of image regions.
  • At least one of the plurality of video frames includes an object in an image region.
  • the method also includes providing a plurality of background classifications that correspond to the background features in the scene and assigning one of the classifications to the one of the image regions based at least in part on a location of the object relative to the image regions.
  • the method includes assigning more than one background classification to an image region.
  • the background classifications include floor, obstruction, and portal.
  • assigning of a background classification to an image region includes comparing a value associated with the image region to a floor threshold, and obstruction threshold and/or a portal threshold.
  • the method of video analysis also includes determining whether an object in a video frame has newly appeared in the video frame and determining in which image regions a newly appeared object is present. In one embodiment, the method includes determining whether an object has newly disappeared from a video frame and determining in which image regions any newly disappeared objects are present. In another embodiment, assigning a classification to an image region depends on the number of times an image has newly appeared in or newly disappeared from an image region. [0018] In a further embodiment, the method includes determining whether to track an object based, at least in part, on the background classification assigned to at least one of the image regions of the video frame in which the object is present.
  • the object in a video frame includes a boundary, which can be a lower boundary, a top boundary, or a side boundary.
  • the method includes the step of determining a boundary region that includes the object's boundary.
  • the method also includes the step of determining whether to track an object based at least in part on the image region in which at least one boundary region is included relative to the background classification of the boundary region or another image region.
  • the method also includes using the size of the object as a factor in determining whether to track an object.
  • the method includes selecting a video frame containing an object and determining whether the object appears in other video frames based, at least in part, on the background classifications assigned to one of the image regions. In one embodiment, the method includes determining whether an object in a video frame appears in another video frame (earlier or a later) based, at least, in part on the background classification assigned to one of the image regions.
  • the invention in another aspect, relates to a computerized method of video analysis that includes receiving image data for a plurality of video frames depicting a scene, wherein at least one video frame contains an object, and determining a vanishing point for that frame based, at least in part, on characteristics of the object and the vanishing point of other frames.
  • the vanishing point determined is a vertical vanishing point.
  • the vertical vanishing point can be determined based on the vertical vanishing points of other frames and characteristics of the object in the video frame.
  • determining the vertical vanishing point is also based on the major axis and the centroid of the object in the video frame.
  • the object being used to determine the vanishing point has a boundary comprising a plurality of pixels.
  • the method includes the step of selecting one of the boundary pixels based at least in part on the vanishing point of the frame.
  • the selected pixel corresponds to the bottom of the boundary, and its selection is based on the vertical vanishing point of the video frame.
  • the invention in a third aspect, relates to a computerized method for video analysis that includes receiving image data for a plurality of video frames depicting a scene that includes at least one of several background features. Each video frame includes a plurality of image regions and at least one video frame has within it an object with a lower boundary.
  • the method further includes providing background classifications corresponding to the background features in the scene, determining a vertical vanishing point for at least one of the video frames that includes an object based on characteristics of the object and on the vertical vanishing point of other video frames, and determining the lower boundary of the object.
  • the method also includes the step of assigning one of the background classifications to at least one image region based at least in part on the location of the lower boundary relative to the location of the image region.
  • the method also includes counting the number of frames in which a lower boundary of an object is present in each object region. In another embodiment, the method also includes counting the number of frames in which an object is present in each image region. In still another embodiment, the step of assigning a background classification to one of the image regions also includes comparing the number of frames in which a lower boundary of an object is present in an image region to a floor threshold and/or comparing the number of times an object is present in an image region to an obstruction threshold. [0025] In another aspect, the invention relates to a video analysis system.
  • the system includes a means for receiving image data for a plurality of video frames that depict a scene.
  • the video frames include at least one of a plurality of background features.
  • Each video frame includes a plurality of image regions, and at least one video frame has an object in at least one image region.
  • the system also includes a means for providing a plurality of background classifications.
  • the system further includes a means for assigning one of the background classifications to at least one of the image regions based at least in part on a location of the object relative to the image regions.
  • the invention in another aspect, relates to a video analysis system that includes a means for receiving image data for a plurality of video frames that depict a scene. At least one video frame has an object within the frame. The system also includes a means for determining a vanishing point for the frame based at least in part on one or more characteristics of the object and the vanishing point of other frames.
  • Fig.l is a block diagram of an illustrative overall computer-assisted surveillance ("CAS") system utilizing one aspect of the invention.
  • CAS computer-assisted surveillance
  • FIG. 2 is a high-level block diagram of a sample embodiment CAS computer.
  • Fig. 3 is a schematic drawing of a video frame including classified image regions according to one embodiment of the invention.
  • Fig. 4 is a flow chart illustrating a background classification method according to one embodiment of the invention.
  • Fig. 5 A is a schematic depiction of a video frame depicting a scene having an object and a portal.
  • Fig. 5B is a schematic depiction of the subsequent video frame to the video frame of Fig. 5A no longer depicting the object.
  • Fig. 5C is a schematic depiction of the background classification analysis step initiated based on the video frames of Figs. 5 A and 5B according to one embodiment of the invention.
  • Fig. 6 A is a video frame demonstrating a first vertical vanishing point in a first scene.
  • Fig. 6B is a video frame demonstrating a second vertical vanishing point in a second scene.
  • Fig. 6C is a video frame demonstrating a third vertical vanishing point in a third scene.
  • Fig. 7 depicts a flow chart illustrating a method of determining a vertical vanishing point for a frame according to one embodiment of the invention.
  • Fig. 8 is an illustrative video frame to be analyzed according to the method of Fig. 7.
  • Fig. 9A is a schematic depiction of a scene having two objects and a piece of furniture.
  • Fig. 9B is a schematic depiction of the background classification analysis step initiated based on the video frame of Figs. 7A.
  • Fig. 10 is a flow chart of a method for determining whether to track an object in a video frame based at least in part on the classification of image regions in the video frame.
  • Fig. 11 A is a three dimensional model of a simulated monitored environment.
  • Fig. 1 IB is a sample video frame analyzed by a CAS computer in a simulation of one embodiment of the invention.
  • Fig. 11C is a sample video frame analyzed by a CAS computer in a simulation of one embodiment of the invention.
  • Fig. 12A is a representation of the status of a floor matrix after the analysis of one hundred simulated video frames according to one embodiment of the invention.
  • Fig. 12B is a representation of the status of a portal matrix after the analysis of one hundred simulated video frames according to one embodiment of the invention.
  • Fig. 12C is a representation of the status of an obstruction matrix after the analysis of one hundred simulated video frames according to one embodiment of the invention.
  • Fig. 13 A is a representation of the status of the floor matrix after the analysis of one thousand simulated video frames according to one embodiment of the invention.
  • Fig. 13B is a representation of the status of the portal matrix after the analysis of one thousand simulated video frames according to one embodiment of the invention.
  • Fig. 13C is a representation of the status of the obstruction matrix after the analysis of one thousand simulated video frames according to one embodiment of the invention.
  • Fig. 14A is a representation of the status of the floor matrix after the analysis often thousand simulated video frames according to one embodiment of the invention.
  • Fig. 14B is a representation of the status of the obstruction matrix after the analysis often thousand simulated video frames according to one embodiment of the invention.
  • Fig. 14C is a representation of the status of the obstruction matrix after the analysis often thousand simulated video frames according to one embodiment of the invention.
  • cameras capture image data that depicts the interaction of people and things in a monitored environment.
  • Types of cameras include analog video cameras, digital video cameras, or any device that can generate image data.
  • the word "camera,” is used as a generic term that encompasses any sensor that can output video data.
  • the CAS system observes a monitored environment through a number of input sensors although its primary sources of information are video cameras.
  • the majority of CCTV installations use common visible-light video cameras.
  • the CAS system employs advanced video analysis algorithms for the extraction of information from analog NTSC or PAL video. These algorithms, however, are not limited to the visible light spectrum; they can also be applied to infrared video or even imagery from radar or sonar installations if available.
  • FIG. 1 shows an illustrative computer-assisted surveillance ("CAS") system 100.
  • a plurality of cameras or other image input devices 102 provide image inputs to a computer 104 programmed to provide image analysis.
  • CAS computer 104 can include a display 106 providing a graphical user interface for setup, control and display.
  • CAS computer 104 can also include one or more user input devices (not shown) such as keyboards, mice, etc. to allow users to input control signals.
  • CAS computer 104 performs advanced image processing including image feature extraction and tracking.
  • CAS computer 104 can automatically detect objects and activity and can generate warning and other information that can be transmitted over a digital communications network or other interface 108.
  • CAS computer 104 also uses interface 108 to retrieve data, such as previously recorded video stored on recorder 112 or information stored on other computers.
  • CAS computer 104 provides the outputs of the various cameras 102 to a multiplexer 110 for recording, typically continuous or stop-frame, by recorder 112 and for display on one or more displays 114 via a switcher 1 16.
  • An additional user interface (e.g., provided by another computer 118 and user input including, for example, a joystick 120) can be used to allow an operator to control switcher 116 to select images to view and to control other parts of system 100 including CAS computer 104.
  • Mutiplexer 110 and/or switcher 116 can respond to external alarms that occur when certain types of activity have been automatically detected (e.g., an alarm generated by a motion sensor) and record or display video appropriately. These alarms can also be generated by CAS computer 104 based on detected activities in the video streams.
  • the illustrative CAS Computer 104 system integrates seamlessly into any existing security infrastructure.
  • the illustrative embodiment CAS system 100 is compatible with, for example, legacy analog video sources, in addition to newer digital video sources such as USB, Fire Wire, or IP cameras on wired or wireless networks.
  • the CAS computer 104 acts as a passive repeater of its input signals, so that in the unlikely event of a CAS computer 104 failure, the remainder of the security infrastructure continues to function without the CAS computer 104.
  • video cameras 102 are the typical primary sensors for the CAS system 100
  • the system can also accommodate other commonly-used sensors, such as motion detectors, smoke detectors, spill detectors, microphones, point-of-sale (POS) recordings, electronic article surveillance (EAS) systems, and access control systems.
  • POS point-of-sale
  • EAS electronic article surveillance
  • the illustrative CAS system 100 combines information from these sensors with the video analysis results to provide an even richer description of activities in the world. For example, POS information may be used with video images to verify that a customer purchased a particular product.
  • Fig. 2 shows a high-level block diagram of an illustrative CAS computer 104.
  • the system components are grouped into two main classes: single-view processing blocks 202 (SVPs) and multi-view processing blocks 204 (MVPs).
  • SVPs single-view processing blocks 202
  • MVPs multi-view processing blocks 204
  • Each image input source is attached to a SVP 202.
  • Image input sources include cameras 102 as well as a variety of storage devices including, for example, computer disks, VHS tapes, and digital videotapes.
  • Each SVP 202 typically performs video processing tasks that require only a single video stream.
  • the outputs of the SVP 202 are connected to a MVP 204 that processes multiple video streams at once.
  • the single-view processing components 202 and the multi-view processing components 204 typically analyze data as a series of video frames depicting a scene.
  • image data is analyzed directly from a camera.
  • the analyzed image data can originate from a storage device.
  • Some cameras and video storage devices create and store image data on a frame-by-frame basis.
  • Other storage systems may only store video frame updates, i.e. detected changes to the scene.
  • the CAS computer 104 constructs a video frame from stored image data that may be stored in a variety of devices and formats.
  • a video frame is composed of a plurality of pixels.
  • the number of pixels in a video frame typically depends on, among other factors, the resolution of the camera generating the video frame, the display on which the video frame is presented, and the capacity of the storage device on which the video frames are stored.
  • Analysis of a video frame can be conducted either at the pixel level or by analyzing groups of pixels depending on the processing power available and the level of precision desired.
  • a pixel or group of pixels to be analyzed is referred to herein as an "image region.”
  • the SVPs 202 of the CAS computer 104 classify the image regions of video frames according to background image features. Identifying background features corresponding to image regions, classifying the regions based upon those background features, and utilizing the information that such classification provides can aid the tracking process.
  • the illustrative embodiment stores information about image regions of a series of video frames in a plurality of two-dimensional background feature matrices, one matrix for each background feature to be determined. For example, one embodiment maintains a floor matrix, an obstruction matrix, and portal matrix. For purposes of populating the matrices, image regions of a video frame are divided into rows and columns. Image regions are referred to by their location ⁇ X J , y j ⁇ , in the background feature matrices.
  • a group of image regions is indicated by a range of row numbers followed by a range of column numbers, for example ⁇ x ⁇ -Xi 2 , yji— Vj 2 ⁇ -
  • the x dimension of the matrix indicates the column number of an image region
  • the y dimension indicates the row number of the image region.
  • one matrix is maintained for all background features.
  • the data is stored in a three-dimensional matrix with the third dimension corresponding to the feature being considered.
  • the image region size i.e., the number of pixels that make up an image region, used for each background feature matrix may differ depending on the background feature analyzed. For example, floor regions may be classified with higher granularity than portal regions.
  • all matrices use the same image region size and that image region ⁇ X J , y j ⁇ in one background feature matrix refers to the same image region ⁇ x,, yj ⁇ in all other background feature matrices.
  • the CAS computer 104 populates the background feature matrices with information relating to each image region by analyzing the position of objects from video frame to video frame.
  • FIG. 3 is a schematic drawing of a video frame 300 depicting a sample scene.
  • the video frame 300 has been overlaid with gridlines showing the boundaries of image regions for illustrative purposes.
  • the sample video frame includes image regions that are classified as floor (e.g., image region ⁇ 7,20 ⁇ ), obstruction (e.g., image region ⁇ 17, 14 ⁇ ), and portal (e.g., image region ⁇ 14,12 ⁇ ).
  • Image regions classified as floor are indicated with the letter "F”
  • image regions classified as obstructions are indicated with the letter "O”
  • image regions classified as portal are indicated with the letter "P.”
  • an image region can be assigned more than one classification (not shown).
  • Image region ⁇ 7, 20 ⁇ is classified as a floor region.
  • Knowledge of floor pixels can be used, for example, to classify objects as interesting or non-interesting.
  • the system is intended to track human activity. As humans usually have at least one foot on the floor as they move, objects detected that are not at least partially present in or near an image region classified as a floor (e.g., a bird) can be ignored if they are not considered interesting objects.
  • an image region is classified as floor as long as bottoms of objects tend to be present in or located near the image region.
  • image region ⁇ 9, 15 ⁇ representing the top of a piece of furniture
  • image region ⁇ 9, 15 ⁇ representing the top of a piece of furniture
  • the piece obstructs the camera's view of the lower portion of the object.
  • the detected bottoms of such objects will be determined to be at the top of the piece of furniture, leading to the classification of the image region as a floor.
  • Image region ⁇ 17, 14 ⁇ is classified as an obstruction region.
  • Obstructions tend to at least partially block the view of objects behind the obstruction from the camera.
  • Typical obstructions include pieces of furniture, walls, and architectural features, such as columns. Knowing that an object is located near an image region classified as obstruction can be useful in determining whether part of an object is occluded from view.
  • Objects may be analyzed at least partially based on their size. If part of an object is occluded, such analysis may be of limited usefulness.
  • Image region ⁇ 14, 12 ⁇ is classified as a portal region.
  • Objects tend to first appear in an image frame in portal regions and objects tend to last appear in an image frame in portal regions. For example, a person usually enters or leaves a scene through a door.
  • the image regions that include the door are classified as portal regions.
  • surrounding image regions may be classified as portals. For example, if people frequently walk behind a large column and are fully hidden from the camera's view, the image regions that include the sides of the column are classified as portal regions (see, e.g., image region ⁇ 19, 11 ⁇ ).
  • Typical portal regions may also include stairwells, elevators, the edges of a camera's field-of-view, and image regions surrounding large pieces of furniture or large architectural features.
  • Knowledge of portals can allow a tracking system to account for the appearance of new objects or the disappearance of old objects. For example, if a scene includes one person far from a portal image region, and a second person subsequently enters the scene at or near the portal image region, a CAS computer 104 is less likely to confuse the first person with the second person if the CAS computer 104 can determine that the second person is likely to be new to the scene since he/she appeared at or near a portal region.
  • Classification of a given image region is based on comparing information stored in the background feature matrices for the image region to a threshold.
  • the CAS computer 104 has a threshold for each background feature. For example, if the value stored the floor matrix for image region ⁇ X J , y j ⁇ exceeds the floor threshold, then image region ⁇ xj, yj ⁇ is classified as a floor.
  • classification depends on comparing the value stored for an image region from two or more background feature matrices.
  • an image region is classified as an obstruction region if the value stored in the obstruction matrix for that image region exceeds the obstruction threshold, but the value of the floor matrix for that image region does not exceed the floor threshold.
  • an image region is classified as a floor region and not an obstruction region if the value of the floor and obstruction matrices for that image region both exceed their respective thresholds. Thresholds are discussed below in further detail.
  • Figure 4 is a flow chart depicting a background classification method 400 according to one embodiment of the invention. To understand the process of maintaining the background feature matrices, Figure 4 will be described in conjunction with Figures 5-8.
  • the CAS computer 104 utilizing the background classification method 400 maintains an object list that includes information about objects that are present in each video frame.
  • the object list may contain identifying information for each object in the video frame, including, for example, the location, color, and size of each object as it appears in the current video frame and as it had appeared in previous video frames.
  • the object remains in the object list until removed.
  • the object list and the background feature matrices are initialized (step 402).
  • the object list is cleared and the data stored for the image regions in the background feature matrices are reset to zero.
  • the background feature matrices store counts of occurrences related to each background feature.
  • the CAS computer 104 then analyzes a series of video frames, one frame at a time. For each video frame, the CAS Computer 104 identifies the objects in the video frame and labels each identified object as unprocessed (step 404). The CAS computer 104 identifies objects using one of several methods commonly known in the art (e.g., background subtraction). In step 406, the CAS computer 104 determines whether there are any unprocessed objects in the object list for the video frame. Unless the video frame does not include any objects or all objects have already been analyzed for the video frame, the CAS computer 104 selects one of the unprocessed objects in the object list for analysis (step 408). After selecting an object (step 408) the CAS computer 104 determines whether the selected object is newly appeared in the scene depicted in the video frame (step 410), i.e. was the object not listed in the object list for the preceding video frame.
  • the CAS computer 104 determines whether the selected object is newly appeared in the scene depicted in the video frame
  • Figure 5A is a video frame 500 depicting a scene at time t.
  • Figure 5B is a video frame 500' depicting the same scene at time t+1, i.e. it is the immediately subsequent video frame captured after the video frame 500 of Figure 5 A.
  • Video frames 500 and 500' are overlaid with gridlines depicting boundaries of image regions for illustrative purposes.
  • the video frame 500 includes a door 502 and no objects.
  • the video frame 500' includes the door 502 and an object 504 standing in the door. Since the object 504 was not in video frame 500, it is not listed in the object list for video frame, and therefore the CAS computer 104 considers the object 504 to be newly appeared.
  • FIG. 5C is a schematic drawing 500" of the steps the CAS computer 104 takes to update the portal matrix based on analyzing video frames 500 and 500'.
  • Video frame 500" is overlaid with gridlines depicting boundaries of image regions for illustrative purposes.
  • the CAS computer 104 increments the count stored for each image region in the portal matrix in which the newly appeared object appears.
  • the newly appeared object appears in image regions ⁇ 5-6, 4— 7 ⁇ .
  • the CAS computer 104 increments the count stored for each image region ⁇ 5-6, 4—7 ⁇ in the portal matrix as indicated in the schematic drawing 500".
  • the CAS computer 104 also adds the object to the object list (step 414).
  • the CAS computer 104 also updates the portal matrix in response to determining that an object is newly disappeared. For example, assume that video frames 500 and 500' were captured and analyzed in reverse order. In video frame 500' at time t, the video frame 500' includes an object 504 standing in the door 502. In video frame 500, the immediately subsequent video frame, at time t+1, the video frame 500 no longer includes the object 504. Since no objects are included in video frame 500, in step 406, when the CAS computer 104 determines whether there are any unprocessed objects, the CAS computer 104 does not locate any unprocessed objects, and proceeds to determine if there are any newly disappeared objects (step 416).
  • the CAS computer 104 analyzes the object list to determine whether the preceding frame 500' included any objects that the current video frame 500 does not (step 418). Upon detection of the object 504 listed in the object list of the preceding frame, the CAS computer 104 updates the portal matrix (step 420) in the same fashion as described above and as depicted in the schematic drawing 500". The object 504 is then removed from the object list (step 422). In one embodiment, if the CAS computer 104 detects a newly disappeared object, the CAS computer 104 determines if the object was present in or located near at least one image region from the previous frame that is classified as portal (see below). If the newly disappeared object was not in or near a portal image region, the CAS computer 104 keeps the object in the object list for at least one additional frame to take into account the possibility of the CAS computer 104 incorrectly missing the object in the video frame.
  • the CAS computer 104 updates the floor matrix (step 424). To do so, the CAS computer 104 first determines the bottom or lower boundary of the object and then increments the count stored in the floor matrix for one or more image regions based on the location of the bottom of the object.
  • Objects have at least one boundary.
  • boundaries are lines composed of pixels that could be traced around the exterior of the object.
  • boundaries are boundary regions that surround particular pixels that make up the object.
  • boundaries are also ascribed labels . For example, if one were to draw a real world- vertical axis (i.e., up and down) through an object in a video frame, a portion of the video frame surrounding a pixel of the object that is furthest along the real world- vertical axis in the down direction is a bottom boundary region.
  • the bottom boundary region is a single pixel. In another embodiments, the bottom boundary region is the image region that includes such a pixel.
  • a set of parallel lines intersects at a common point called the vanishing point. Every set of parallel lines intersects at a different vanishing point.
  • the location of vertical vanishing point determines the up direction, and consequently the up direction may vary at different pixels in the image.
  • Figures 6A-6C are three video frames 600, 600', and 600", respectively, each demonstrating one of three distinct possible vertical vanishing points.
  • Video frame 600 is an example of a video frame in which the vertical vanishing point 602 conceptually lies off of the screen.
  • Video frame 600' is an example of a video frame in which the vertical vanishing point 602' is at infinity
  • video frame 600" is an example of a video frame in which the vertical vanishing 602" point lies within the video frame 600" itself.
  • Video frames 600, 600' and 600" each include several objects 604, 604' and 604".
  • Each object 604, 604' and 604" has a vertical axis 606, 606', and 606" that is an extension of a line that conceptually connects the centroids 608, 608' and 608" of the objects 604, 604', and 6004" with the vertical vanishing point 602, 602', and 602".
  • the vertical axes 606 for all objects 604 are generally the same direction, though it varies somewhat depending upon the video frame-horizontal (the direction of pixel rows) position of the object 604.
  • the vertical axes 606' of the objects 604' are parallel, i.e. they share the same up direction, corresponding to the direction of pixel columns.
  • the vertical axes 606" of the objects 604" in video frame 600" vary greatly.
  • the up direction is towards the left.
  • the up direction is generally to the right.
  • up is generally away from the center of the video frame 600".
  • An overhead camera for example, can create video frame 600".
  • Figure 7 is a flow chart of an illustrative computerized method of determining the vertical vanishing point 700 for a series of video frames according to one embodiment of the invention.
  • the CAS computer 104 calculates the vertical vanishing point based on the orientation of a plurality of objects in the series of video frames.
  • Figure 7 should be viewed in conjunction with Figure 8 for an understanding of the vertical vanishing point calculation method 700.
  • Figure 8 is one illustrative video frame to be analyzed according to the method 700.
  • the vanishing point is continuously updated with each video frame; while in other embodiments, the vanishing point is updated periodically.
  • the CAS computer 104 For each video frame analyzed, the CAS computer 104 identifies the objects 802, 804 and 806 that are included in the video frame 800 (step 702). The CAS computer 104 then selects an object 802 to analyze (step 704). For each object, the CAS computer 104 carries out the following analytical steps.
  • the CAS computer 104 analyzes the selected object 802 to determine if the object 802 appears to be a person (step 706). To do so, the CAS computer 104 computes the major axis 808 and minor axis 810 of the object 802. This computation is a standard computer vision procedure described in introductory textbooks, such as Robot Vision by Berthold Horn, and published by M.I.T. Press in 1986. Intuitively, the major axis 808 points along the "long" dimension of an object 802 and the minor axis 810 points along the "short" dimension. In the case of people, the major axis 808 generally points in the head-to-foot dimension (i.e., the vertical direction) and the minor axis 810 is perpendicular to that.
  • the major axis 808 and minor axis 810 are about the same size and point in random directions.
  • the CAS computer 104 computes the ratio of the major axis to the minor axis. If this ratio is big enough, for example, in one illustrative embodiment, greater than 1.5, then the object is considered for further processing.
  • the CAS computer 104 determines whether the orientation of the object 802 substantially comports with the current vertical vanishing point estimate 816. In one illustrative embodiment, the CAS computer 104 computes a line 812 that should intersect with the vertical vanishing point 816. This line 812 has the slope of the major axis 808 and contains the centroid 814 of the object. The centroid 814 is typically the point at which the object's major axis 808 and minor axis 810 intersect. In one embodiment, the CAS computer 104 then computes the perpendicular distance 817 between this line 812 and the vertical vanishing point 816.
  • the object is processed further.
  • the CAS computer 104 compares a line 818 that passes through both the vertical vanishing point 816 and the centroid 814 of the object 802. If the angle between of the aforementioned lines, 812 and 818, is large, for example if the object 802 is in a prone position or is in the process of falling, the CAS computer 104 does not use selected object 802 as a new data point in updating the vertical vanishing point 816. If the angle between the two lines is less than a small angle (e.g., 0.1 degrees), then the object is processed further. If the CAS computer 104 has selected the object for further processing, the CAS computer 104 incorporates the object into its calculation of a new vertical vanishing point for the video frame (step 710).
  • a small angle e.g., 0.1 degrees
  • Steps 704, 706, 708, and 710 are repeated for each object in the video frame.
  • the entire method 700 then repeats with the next video frame.
  • the CAS computer 104 builds a 3X3 vertical vanishing point matrix, updating the matrix with information from the objects selected for further processing.
  • two-dimensional points are represented by three numbers (so-called homogeneous coordinates).
  • point (x,y,l) corresponds to pixel (x,y) and point (x,y,0) corresponds to a point at infinity.
  • is a weight parameter that varies from 0 to 1. An ⁇ value closer to 1 weights
  • the CAS computer 104 determines the bottom or bottom boundary of an object, after which the floor matrix can be updated accordingly (step 424).
  • Figure 9 A is a video frame 900 depicting a first object 902 and a second object 904 in a scene.
  • the video frame 900 is overlaid with gridlines indicating the boundaries of image regions.
  • the second object 904 is standing behind a piece of furniture 906.
  • Figure 9B is a schematic drawing 900', including demonstrative image region boundaries, of the steps the CAS computer 104 takes to update the floor and obstruction matrices based on analyzing video frame 900. It is assumed that the vertical vanishing point in the video frame 900 is at infinity and that the vertical axis 901 corresponds to the image region columns, i.e. towards the top of the page is up and towards the bottom of the page is down.
  • the bottom boundary region for the first object 902 appears in image region ⁇ 6, 9 ⁇ .
  • the count stored in location ⁇ 6, 9 ⁇ of the floor matrix is incremented.
  • the bottom boundary region of the second object 904 appears in image region ⁇ 8, 8 ⁇ . Note that the bottom boundary region of the second object 904 does not correspond to the feet (not shown) of the second object 904. It corresponds however, to the lower most, i.e. furthest in the down direction, portion of the second object 904 that is visible to the camera that is taking the video.
  • the count stored in floor matrix location ⁇ 8, 8 ⁇ is also incremented.
  • additional counts for image regions surrounding the bottom of image may also be incremented.
  • the counts stored in floor matrix locations ⁇ 5-7, 9-10 ⁇ are incremented. Incrementing the counts surrounding regions is useful when the CAS computer 104 is using relatively small image regions, thereby populating the matrix faster.
  • Fig. 9B also indicates the changes to the counts stored in the obstruction matrix that an illustrative embodiment of the CAS computer 104 would make based on the video frame of Fig. 9A (step 426).
  • locations in the obstruction matrix corresponding to an image region in which an object is present, other than the object's bottom boundary region is incremented.
  • Fig. 9B indicates that the counts stored in obstruction matrix locations ⁇ 5-6, 5-8 ⁇ and ⁇ 5,9 ⁇ would be incremented in updating the obstruction matrix (step 426) for the first object 902.
  • the counts stored in obstruction matrix locations ⁇ 8, 6-7 ⁇ , and ⁇ 9, 7 ⁇ would be incremented in updating the obstruction matrix (step 426) or the second object 904.
  • the obstruction matrix location associated with the object's bottom boundary region is also incremented.
  • the CAS computer 104 determines whether any other objects are present in the video frame that have not been analyzed (step 406). If no additional objects are left to analyze, the video frame determines if there are any newly disappeared objects (step 416) as described above with reference to Figure 5. After all newly disappeared objects are analyzed, the CAS computer 104 advances to the next frame in the series of video frames (step 428), and the process repeats beginning with step 404, by identifying the objects in the new video frame.
  • Figure 10 is a flow chart of a method for determining whether to track an object in a video frame based at least in part on the classification of image regions in the video frame 1000.
  • the CAS computer 104 utilizes the data accumulated in the background feature matrices to classify image regions.
  • the CAS computer 104 In determining whether to track objects in a video frame, the CAS computer 104 first selects an object in the video frame (step 1002). The CAS computer 104 then determines whether the selected object is on the floor (step 1004), i.e. is the bottom boundary of the object present in an image region that is classified as floor. Determining the bottom boundary of the object is described above with respect to Figures 6-8. The CAS computer 104 then determines whether the image region in which the bottom boundary is classified as floor. In one embodiment, the CAS computer 104 compares the count stored in the image region's location in the floor matrix to a floor threshold. If the count exceeds the threshold, the image region is classified as floor.
  • the floor threshold is determined dynamically based on the following factors: a pre-established estimated value p f , ranging from 0 to 1, representing the proportion of the matrix elements that are being updated correctly; a pre-established estimated value f f , ranging from 0 to 1, representing the fraction of the image regions of the video frame that should be classified as floor regions; the sum Nf of the counts for all image regions in the floor matrix; and the number of image regions M f in the video frame.
  • the floor threshold F thresh is derived from the following equations: P f N f f f M f
  • each classification has its own threshold such as a floor threshold, an obstruction threshold and a portal threshold.
  • the thresholds for other classifications are determined by the same equation and substituting p 0 , f 0 , M 0 , and N 0 for the obstruction threshold equation and p p , f p , M p , and N p for the portal threshold equation.
  • the CAS computer 104 updates the f value for each classification based on the number of image regions the CAS computer 104 has classified as belonging to each classification.
  • the CAS computer 104 tests the size of the object to determine if the object is large enough to be tracked by the CAS computer 104 (step 1006).
  • the size is an absolute number, for example, the object must include at least 25 pixels.
  • the size might be relative.
  • the CAS computer 104 may compare the length of the major axis of the object with the minor object of the axis as discussed above with respect to determining vertical vanishing points.
  • the CAS computer 104 determines whether the object is touching an obstruction that might be occluding part of the object (step 1008). For example, referring back to Figure 9, the second object 904 may not be large enough to be tracked because part of the second object 904 is occluded by a piece of furniture. In that case, in the illustrative embodiment, the CAS computer 104 determines whether the lower boundary of the object touches an occluding object. To do so, the CAS computer 104 determines the lower boundary of the second object. The CAS computer 104 then determines the classification of the image region below that in which the lower boundary of the object is present. If that region is classified as an obstruction, that embodiment of the CAS computer 104 tracks the object even if it is not located on the floor or if the object appears to be too small to be tracked.
  • the object is discarded (step 1010) and the CAS computer 104 begins analyzing the next object. If the CAS computer 104 determines to track an object, the CAS computer 104 determines whether the object is newly appeared in the scene (step 1012). In one embodiment, the CAS computer 104 maintains an object list that maintains information, including location information, about objects in a scene. Based at least in part on the information stored in the object list, the CAS computer 104 determines whether the object in the video frame being analyzed likely corresponds to an object listed in the object list from the previous video frame. If the object likely corresponds to an object listed in the object list from a previous video frame, the CAS computer 104 continues to track the object (step 1014).
  • the CAS computer 104 determines whether the non-corresponding object is present in or located near an image region classified as portal (step 1016). If the non-corresponding object is not present in or located near an image region classified as portal, the CAS computer 104 adds the object to the object list and begins tracking the object (step 1018). The CAS computer 104 then begins analyzing the next object in the video frame (1002), if any remain to be analyzed.
  • the CAS computer 104 searches for objects that were present in or located near image regions classified as portals in previous frames and in frames captured by other cameras (1020). For example, if the CAS computer 104 were analyzing two connecting rooms, each with its own camera, and the CAS computer 104 determines that a non-corresponding object is present in an image region classified as portal in the first room, the CAS computer 104 would check the object list to determine if, in a video frame captured by the camera monitoring the second room, an object was present in a region classified as portal in the second room. The CAS computer 104 then determines whether any such objects ("found objects") were found (step 1022).
  • the CAS computer 104 determines that the non-corresponding object corresponds to the found object, and the object continues to be tracked as such (1014). If there is no found object, the CAS computer 104 adds the object to the object list and begins to track it as a newly appeared object (step 1018).
  • a simulation of the background classification system can be prepared by designing a three-dimensional model of a typical surveillance environment with person-sized objects programmed to travel through the model according to random trajectories.
  • the CAS computer 104 then monitors the simulation over a time interval collecting image data as if the graphical output produced in the simulation could be attributed to a plurality of video frames.
  • Figs. 11A-C depict three video frames from one such above described simulation of a monitored environment.
  • the monitored environment 1100 is devoid of objects.
  • the monitored environment 1100 includes a door 1102, a piece of furniture 1104, and a pillar 1106.
  • objects 1108 are present (depicted as white rectangular prisms) in the monitored environment.
  • the frame-rate of the simulation is approximately 25 frames per second.
  • Figs. 12A-12C, 13A-13C, and 14A-14C are graphical depictions of the state of the background classification matrices populated during the simulation after the CAS computer 104 has analyzed one hundred frames of image data (approximately 5 seconds), one thousand frames of image data (approximately one minute), and ten thousand frames of image data (approximately 10 minutes), respectively.
  • the shaded region 1200a in Fig. 12A depicts the state of the floor matrix after the analysis of one hundred frames.
  • the shaded region 1200a includes the image regions that have been classified as floor image images.
  • the shaded region 1200b includes the image regions that have been classified as portal image regions.
  • the shaded region 1200c includes the image regions that have been classified as obstruction image regions, but not floor image regions ("non-floor obstruction image regions").
  • the shaded regions 1300a, 1300b, and 1300c in Figures 13A, 13B, and 13C represent the image regions that are classified as floor regions, portal regions and non-floor obstruction regions after the analysis of one thousand video frames.
  • more image regions depicted in Figures 12A-12C more image regions have been classified in Figs. 13A-C.
  • Fig. 13 A most of the floor regions have been identified, including an area on top of the piece of furniture 1104 where objects appear as they move behind the piece of furniture 1104.
  • Fig. 13B the image regions around the door 1 102, on either side of the pillar 1106 and around much of the periphery of the environment 1100 have correctly been identified as portal regions.
  • Fig. 14A-C after ten thousand frames, the classification of image regions, as depicted with the shaded regions 1400a, 1400b and 1400c (floor, portal, and non-floor obstruction regions, respectively), has become more refined, i.e. boundaries of classifications have taken on contoured shapes as opposed to the rougher boundaries noticeable in Figs. 13A- 13C.
  • the classifications depicted in Figs. 14A-C however, have not significantly changed from their state after the analysis of only one thousand frames, demonstrating that the system can (depending on the number and of objects moving within a monitored environment) generate useful results after only a relatively short monitoring time.

Abstract

A computerized method of video analysis including receiving image data for a plurality of video frames depicting a scene that includes at least one of a plurality of background features. Each video frame includes a plurality of image regions and at least one video frame has an object within at least one of the image region. A plurality of background classifications is provided that correspond to one of the background features in the scene. At least one image region is assigned a background classification based at least in part on the location of the object relative to the image region.

Description

METHOD AND APPARATUS FOR COMPUTERIZED IMAGE BACKGROUND ANALYSIS
Technical Field
[0001] The present invention generally relates to video surveillance, and more specifically to a computer-aided surveillance system for analyzing image backgrounds.
Background
[0002] The current heightened sense of security and declining cost of camera equipment have resulted in increased use of closed circuit television (CCTV) surveillance systems. Such systems have the potential to reduce crime, prevent accidents, and generally increase security in a wide variety of environments.
[0003] A simple closed-circuit television system uses a single camera connected to a display device. More complex systems can have multiple cameras and/or multiple displays. One known type of system is the security display in a retail store, which switches periodically between different cameras to provide different views of the store. Higher security installations, such as prisons and military installations, use a bank of video displays each displaying the output of an associated camera. A guard or human attendant constantly watches the various screens looking for suspicious activity.
[0004] More recently, inexpensive digital cameras have become popular for security and other applications. In addition, it is now possible to use a web cam to monitor a remote location. Web cams typically have relatively slow frame rates, but are sufficient for some security applications. Inexpensive cameras that transmit signals wirelessly (e.g., X 10 cameras) and over power cables to remotely located computer or other displays are also used to provide video surveillance.
[0005] As the number of cameras increases, the amount of raw information that needs to be processed and analyzed also increases. Computer technology can be used to alleviate this raw data processing task, resulting in a new breed of information technology device ~ the computer-aided surveillance (CAS) system. Computer-aided surveillance technology has been developed for various applications. For example, the military has used computer-aided image processing to provide automated targeting and other assistance to fighter pilots and other personnel. In addition, computer-aided surveillance has been applied to monitor activity in swimming pools.
[0006] A CAS system automatically monitors objects (e.g., people, inventory, etc.) as they appear in series of surveillance video frames. One particularly useful monitoring task is tracking the movements of objects in a monitored area. Methods for tracking objects, such as people, moving through an image are known in the art. To achieve more accurate tracking information, the CAS system can utilize knowledge about the basic elements of the images depicted in the series of surveillance video frames.
[0007] On a macroscopic level, a video surveillance frame depicts an image of a scene in which people and things move and interact. On a microscopic level, a video frame is composed of a plurality of pixels, often arranged in a grid-like fashion. The number of pixels in an image depends on several factors including the resolution of the camera generating the image, the display on which the image is presented, the capacity of the storage device on which the images are stored, etc. Analysis of a video frame can be conducted either at the pixel level or at the (pixel) group level depending on the processing capability and the desired level of precision. A pixel or group of pixels being analyzed is referred to herein as an "image region." [0008] Image regions can be categorized as depicting part of the background of the scene or as depicting a foreground object. In general, the background remains relatively static in each frame. However, objects are depicted in different image regions in different frames. Several methods for separating objects in a video frame from the background of the frame, referred to as object extraction, are known in the art. A common approach is to use a technique called "background subtraction." Of course, other techniques can be used. The locations of the objects are typically recorded in a list that is associated with the video frame.
[0009] To track an object from frame to frame, a tracking method determines a correspondence between extracted objects in the current frame and extracted objects in the previous frame. This correspondence can be determined, for example, by using a predictive tracking method. The CAS system predicts the location of an object in the current frame based on the known locations of the object in previous frames. Subsequently, the predicted object location is compared to the actual object location to establish correspondence. Such a prediction is typically based on an algorithm that predicts likely object movement. For example, it can be assumed that objects move with constant velocity. More sophisticated techniques can, for example, verify that the colors of the objects match before determining a correspondence.
[0010] While conceptually simple, a robust tracking system faces many difficulties. Changes in scene lighting can affect the quality of object extraction, causing foreground elements to be misshapen or omitted completely. Object occlusions can cause objects to disappear or merge together, leading to difficulties in correspondence between frames. The tracked objects can change shape or color over time, preventing correspondence even though the objects were properly extracted. [0011] In addition, even under ideal conditions, single-view tracking systems invariably lose track of monitored objects that leave the field-of-view of the camera. When multiple cameras are available, as in many close-captioned television systems, it is theoretically possible to reacquire the target when it appears in a different camera. This ability to perform automatic "sensor hand-off is of significant practical interest. Current laboratory solutions require geometrically calibrated cameras with overlapping fields-of-view, conditions that are not readily achieved in typical CCTV installation.
Summary of the Invention
[0012] It is an object of the present invention to automate the classification of image regions in a series of video frames so that information about background features of a scene can accurately be utilized in aiding CAS systems. Identifying background features corresponding to image regions, classifying the regions based upon those background features, and utilizing the information that such classification provides can aid the object tracking process.
[0013] Examples of background features can include: (i) frequently being immediately "below" or at the bottom of objects in video frames, (ii) frequently having objects first appear in a series of video frames in the image region, (iii) having objects frequently leave a series of video frames from the image region, (iv) being capable of occluding other objects, etc. Image regions having one or more of these features can be classified as being a floor, a portal, or an obstruction.
[0014] A human operator can assign these classifications to image regions based on these or other features. By using a user interface such as that used in a digital paint program, an operator can "paint" classifications onto a video image in areas that correspond to floors, obstructions, and portals. These classifications may not be mutually exclusive, so the different classifications can be represented with different "layers." While painting these pixels might be trivial, it is a time consuming and error-prone task. In a surveillance system with dozens or hundreds of cameras, the human effort involved would be prohibitive, which would make a CAS system that uses this information difficult to deploy in large-scale installations.
[0015] In one aspect, the invention relates to a computer method of video analysis. The method includes receiving image data for a plurality of video frames that depict a scene. The scene exhibits at least one of a plurality of background features and each video frame is composed of a plurality of image regions. At least one of the plurality of video frames includes an object in an image region. The method also includes providing a plurality of background classifications that correspond to the background features in the scene and assigning one of the classifications to the one of the image regions based at least in part on a location of the object relative to the image regions. In one embodiment, the method includes assigning more than one background classification to an image region.
[0016] In one embodiment, the background classifications include floor, obstruction, and portal. In a further embodiment, assigning of a background classification to an image region includes comparing a value associated with the image region to a floor threshold, and obstruction threshold and/or a portal threshold.
[0017] In another embodiment, the method of video analysis also includes determining whether an object in a video frame has newly appeared in the video frame and determining in which image regions a newly appeared object is present. In one embodiment, the method includes determining whether an object has newly disappeared from a video frame and determining in which image regions any newly disappeared objects are present. In another embodiment, assigning a classification to an image region depends on the number of times an image has newly appeared in or newly disappeared from an image region. [0018] In a further embodiment, the method includes determining whether to track an object based, at least in part, on the background classification assigned to at least one of the image regions of the video frame in which the object is present.
[0019] In another embodiment, the object in a video frame includes a boundary, which can be a lower boundary, a top boundary, or a side boundary. The method includes the step of determining a boundary region that includes the object's boundary. In a further embodiment, the method also includes the step of determining whether to track an object based at least in part on the image region in which at least one boundary region is included relative to the background classification of the boundary region or another image region. In a further embodiment, the method also includes using the size of the object as a factor in determining whether to track an object.
[0020] In still another embodiment, the method includes selecting a video frame containing an object and determining whether the object appears in other video frames based, at least in part, on the background classifications assigned to one of the image regions. In one embodiment, the method includes determining whether an object in a video frame appears in another video frame (earlier or a later) based, at least, in part on the background classification assigned to one of the image regions.
[0021] In another aspect, the invention relates to a computerized method of video analysis that includes receiving image data for a plurality of video frames depicting a scene, wherein at least one video frame contains an object, and determining a vanishing point for that frame based, at least in part, on characteristics of the object and the vanishing point of other frames. In one embodiment, the vanishing point determined is a vertical vanishing point. The vertical vanishing point can be determined based on the vertical vanishing points of other frames and characteristics of the object in the video frame. In another embodiment, determining the vertical vanishing point is also based on the major axis and the centroid of the object in the video frame.
[0022] In a further embodiment, the object being used to determine the vanishing point has a boundary comprising a plurality of pixels. The method includes the step of selecting one of the boundary pixels based at least in part on the vanishing point of the frame. In another embodiment, the selected pixel corresponds to the bottom of the boundary, and its selection is based on the vertical vanishing point of the video frame.
[0023] In a third aspect, the invention relates to a computerized method for video analysis that includes receiving image data for a plurality of video frames depicting a scene that includes at least one of several background features. Each video frame includes a plurality of image regions and at least one video frame has within it an object with a lower boundary. The method further includes providing background classifications corresponding to the background features in the scene, determining a vertical vanishing point for at least one of the video frames that includes an object based on characteristics of the object and on the vertical vanishing point of other video frames, and determining the lower boundary of the object. The method also includes the step of assigning one of the background classifications to at least one image region based at least in part on the location of the lower boundary relative to the location of the image region.
[0024] In one embodiment, the method also includes counting the number of frames in which a lower boundary of an object is present in each object region. In another embodiment, the method also includes counting the number of frames in which an object is present in each image region. In still another embodiment, the step of assigning a background classification to one of the image regions also includes comparing the number of frames in which a lower boundary of an object is present in an image region to a floor threshold and/or comparing the number of times an object is present in an image region to an obstruction threshold. [0025] In another aspect, the invention relates to a video analysis system. The system includes a means for receiving image data for a plurality of video frames that depict a scene. The video frames include at least one of a plurality of background features. Each video frame includes a plurality of image regions, and at least one video frame has an object in at least one image region. The system also includes a means for providing a plurality of background classifications.
Each classification corresponds to one of the background features in the scene. The system further includes a means for assigning one of the background classifications to at least one of the image regions based at least in part on a location of the object relative to the image regions.
[0026] In another aspect, the invention relates to a video analysis system that includes a means for receiving image data for a plurality of video frames that depict a scene. At least one video frame has an object within the frame. The system also includes a means for determining a vanishing point for the frame based at least in part on one or more characteristics of the object and the vanishing point of other frames.
Brief Description of the Drawings
[0027] The foregoing discussion will be understood more readily from the following detailed description of the invention, when taken in conjunction with the accompanying drawings:
[0028] Fig.l is a block diagram of an illustrative overall computer-assisted surveillance ("CAS") system utilizing one aspect of the invention.
[0029] Fig. 2 is a high-level block diagram of a sample embodiment CAS computer.
[0030] Fig. 3 is a schematic drawing of a video frame including classified image regions according to one embodiment of the invention.
[0031] Fig. 4 is a flow chart illustrating a background classification method according to one embodiment of the invention. [0032] Fig. 5 A is a schematic depiction of a video frame depicting a scene having an object and a portal.
[0033] Fig. 5B is a schematic depiction of the subsequent video frame to the video frame of Fig. 5A no longer depicting the object.
[0034] Fig. 5C is a schematic depiction of the background classification analysis step initiated based on the video frames of Figs. 5 A and 5B according to one embodiment of the invention.
[0035] Fig. 6 A is a video frame demonstrating a first vertical vanishing point in a first scene.
[0036] Fig. 6B is a video frame demonstrating a second vertical vanishing point in a second scene.
[0037] Fig. 6C is a video frame demonstrating a third vertical vanishing point in a third scene.
[0038] Fig. 7 depicts a flow chart illustrating a method of determining a vertical vanishing point for a frame according to one embodiment of the invention.
[0039] Fig. 8 is an illustrative video frame to be analyzed according to the method of Fig. 7.
[0040] Fig. 9A is a schematic depiction of a scene having two objects and a piece of furniture.
[0041] Fig. 9B is a schematic depiction of the background classification analysis step initiated based on the video frame of Figs. 7A.
[0042] Fig. 10 is a flow chart of a method for determining whether to track an object in a video frame based at least in part on the classification of image regions in the video frame.
[0043] Fig. 11 A is a three dimensional model of a simulated monitored environment.
[0044] Fig. 1 IB is a sample video frame analyzed by a CAS computer in a simulation of one embodiment of the invention. [0045] Fig. 11C is a sample video frame analyzed by a CAS computer in a simulation of one embodiment of the invention.
[0046] Fig. 12A is a representation of the status of a floor matrix after the analysis of one hundred simulated video frames according to one embodiment of the invention.
[0047] Fig. 12B is a representation of the status of a portal matrix after the analysis of one hundred simulated video frames according to one embodiment of the invention.
[0048] Fig. 12C is a representation of the status of an obstruction matrix after the analysis of one hundred simulated video frames according to one embodiment of the invention.
[0049] Fig. 13 A is a representation of the status of the floor matrix after the analysis of one thousand simulated video frames according to one embodiment of the invention.
[0050] Fig. 13B is a representation of the status of the portal matrix after the analysis of one thousand simulated video frames according to one embodiment of the invention.
[0051] Fig. 13C is a representation of the status of the obstruction matrix after the analysis of one thousand simulated video frames according to one embodiment of the invention.
[0052] Fig. 14A is a representation of the status of the floor matrix after the analysis often thousand simulated video frames according to one embodiment of the invention.
[0053] Fig. 14B is a representation of the status of the obstruction matrix after the analysis often thousand simulated video frames according to one embodiment of the invention.
[0054] Fig. 14C is a representation of the status of the obstruction matrix after the analysis often thousand simulated video frames according to one embodiment of the invention. DETAILED DESCRIPTION
[0055] In a surveillance system, cameras capture image data that depicts the interaction of people and things in a monitored environment. Types of cameras include analog video cameras, digital video cameras, or any device that can generate image data. The word "camera," is used as a generic term that encompasses any sensor that can output video data. In one embodiment, the CAS system observes a monitored environment through a number of input sensors although its primary sources of information are video cameras. The majority of CCTV installations use common visible-light video cameras. In such installations, the CAS system employs advanced video analysis algorithms for the extraction of information from analog NTSC or PAL video. These algorithms, however, are not limited to the visible light spectrum; they can also be applied to infrared video or even imagery from radar or sonar installations if available.
[0056] Fig. 1 shows an illustrative computer-assisted surveillance ("CAS") system 100. A plurality of cameras or other image input devices 102 provide image inputs to a computer 104 programmed to provide image analysis. CAS computer 104 can include a display 106 providing a graphical user interface for setup, control and display. CAS computer 104 can also include one or more user input devices (not shown) such as keyboards, mice, etc. to allow users to input control signals.
[0057] CAS computer 104 performs advanced image processing including image feature extraction and tracking. CAS computer 104 can automatically detect objects and activity and can generate warning and other information that can be transmitted over a digital communications network or other interface 108. CAS computer 104 also uses interface 108 to retrieve data, such as previously recorded video stored on recorder 112 or information stored on other computers. CAS computer 104 provides the outputs of the various cameras 102 to a multiplexer 110 for recording, typically continuous or stop-frame, by recorder 112 and for display on one or more displays 114 via a switcher 1 16. An additional user interface (e.g., provided by another computer 118 and user input including, for example, a joystick 120) can be used to allow an operator to control switcher 116 to select images to view and to control other parts of system 100 including CAS computer 104. Mutiplexer 110 and/or switcher 116 can respond to external alarms that occur when certain types of activity have been automatically detected (e.g., an alarm generated by a motion sensor) and record or display video appropriately. These alarms can also be generated by CAS computer 104 based on detected activities in the video streams.
[0058] The illustrative CAS Computer 104 system integrates seamlessly into any existing security infrastructure. The illustrative embodiment CAS system 100 is compatible with, for example, legacy analog video sources, in addition to newer digital video sources such as USB, Fire Wire, or IP cameras on wired or wireless networks. The CAS computer 104 acts as a passive repeater of its input signals, so that in the unlikely event of a CAS computer 104 failure, the remainder of the security infrastructure continues to function without the CAS computer 104.
[0059] While video cameras 102 are the typical primary sensors for the CAS system 100, the system can also accommodate other commonly-used sensors, such as motion detectors, smoke detectors, spill detectors, microphones, point-of-sale (POS) recordings, electronic article surveillance (EAS) systems, and access control systems. The illustrative CAS system 100 combines information from these sensors with the video analysis results to provide an even richer description of activities in the world. For example, POS information may be used with video images to verify that a customer purchased a particular product.
[0060] Fig. 2 shows a high-level block diagram of an illustrative CAS computer 104. For illustrative purposes, the system components are grouped into two main classes: single-view processing blocks 202 (SVPs) and multi-view processing blocks 204 (MVPs). Each image input source is attached to a SVP 202. Image input sources include cameras 102 as well as a variety of storage devices including, for example, computer disks, VHS tapes, and digital videotapes. Each SVP 202 typically performs video processing tasks that require only a single video stream. The outputs of the SVP 202 are connected to a MVP 204 that processes multiple video streams at once.
[0061] The single-view processing components 202 and the multi-view processing components 204 typically analyze data as a series of video frames depicting a scene. In one embodiment, image data is analyzed directly from a camera. In another embodiment, the analyzed image data can originate from a storage device. Some cameras and video storage devices create and store image data on a frame-by-frame basis. Other storage systems may only store video frame updates, i.e. detected changes to the scene. To carry out analysis of image data, the CAS computer 104 constructs a video frame from stored image data that may be stored in a variety of devices and formats.
[0062] A video frame is composed of a plurality of pixels. The number of pixels in a video frame typically depends on, among other factors, the resolution of the camera generating the video frame, the display on which the video frame is presented, and the capacity of the storage device on which the video frames are stored. Analysis of a video frame can be conducted either at the pixel level or by analyzing groups of pixels depending on the processing power available and the level of precision desired. A pixel or group of pixels to be analyzed is referred to herein as an "image region."
[0063] In one illustrative embodiment, the SVPs 202 of the CAS computer 104, classify the image regions of video frames according to background image features. Identifying background features corresponding to image regions, classifying the regions based upon those background features, and utilizing the information that such classification provides can aid the tracking process.
[0064] The illustrative embodiment stores information about image regions of a series of video frames in a plurality of two-dimensional background feature matrices, one matrix for each background feature to be determined. For example, one embodiment maintains a floor matrix, an obstruction matrix, and portal matrix. For purposes of populating the matrices, image regions of a video frame are divided into rows and columns. Image regions are referred to by their location {XJ, yj}, in the background feature matrices. A group of image regions is indicated by a range of row numbers followed by a range of column numbers, for example {xπ-Xi2, yji— Vj2}- The x dimension of the matrix indicates the column number of an image region, and the y dimension indicates the row number of the image region. In other embodiments, one matrix is maintained for all background features. The data is stored in a three-dimensional matrix with the third dimension corresponding to the feature being considered.
[0065] The image region size, i.e., the number of pixels that make up an image region, used for each background feature matrix may differ depending on the background feature analyzed. For example, floor regions may be classified with higher granularity than portal regions. For the illustrative embodiment though, all matrices use the same image region size and that image region {XJ, yj} in one background feature matrix refers to the same image region {x,, yj} in all other background feature matrices. The CAS computer 104 populates the background feature matrices with information relating to each image region by analyzing the position of objects from video frame to video frame.
[0066] The series of video frames depict a scene, and Figure 3 is a schematic drawing of a video frame 300 depicting a sample scene. The video frame 300 has been overlaid with gridlines showing the boundaries of image regions for illustrative purposes. The sample video frame includes image regions that are classified as floor (e.g., image region {7,20}), obstruction (e.g., image region { 17, 14}), and portal (e.g., image region { 14,12}). Image regions classified as floor are indicated with the letter "F," image regions classified as obstructions are indicated with the letter "O," and image regions classified as portal are indicated with the letter "P." In some embodiments, an image region can be assigned more than one classification (not shown).
[0067] Image region {7, 20} is classified as a floor region. Knowledge of floor pixels can be used, for example, to classify objects as interesting or non-interesting. In a typical CAS environment, the system is intended to track human activity. As humans usually have at least one foot on the floor as they move, objects detected that are not at least partially present in or near an image region classified as a floor (e.g., a bird) can be ignored if they are not considered interesting objects.
[0068] In the illustrative embodiment, an image region is classified as floor as long as bottoms of objects tend to be present in or located near the image region. For example, image region {9, 15}, representing the top of a piece of furniture, is classified as floor even though one would not normally refer to the top of a piece of furniture as a floor in common parlance. In this instance, as objects move behind the piece of furniture, the piece obstructs the camera's view of the lower portion of the object. As a result, the detected bottoms of such objects will be determined to be at the top of the piece of furniture, leading to the classification of the image region as a floor.
[0069] Image region {17, 14}, is classified as an obstruction region. Obstructions tend to at least partially block the view of objects behind the obstruction from the camera. Typical obstructions include pieces of furniture, walls, and architectural features, such as columns. Knowing that an object is located near an image region classified as obstruction can be useful in determining whether part of an object is occluded from view. Objects may be analyzed at least partially based on their size. If part of an object is occluded, such analysis may be of limited usefulness.
Knowledge of obstruction image regions can help compensate for such potentialities.
[0070] Image region {14, 12} is classified as a portal region. Objects tend to first appear in an image frame in portal regions and objects tend to last appear in an image frame in portal regions. For example, a person usually enters or leaves a scene through a door. In the illustrative embodiment, the image regions that include the door are classified as portal regions. In other embodiments, surrounding image regions may be classified as portals. For example, if people frequently walk behind a large column and are fully hidden from the camera's view, the image regions that include the sides of the column are classified as portal regions (see, e.g., image region {19, 11 }). Typical portal regions may also include stairwells, elevators, the edges of a camera's field-of-view, and image regions surrounding large pieces of furniture or large architectural features.
[0071] Knowledge of portals can allow a tracking system to account for the appearance of new objects or the disappearance of old objects. For example, if a scene includes one person far from a portal image region, and a second person subsequently enters the scene at or near the portal image region, a CAS computer 104 is less likely to confuse the first person with the second person if the CAS computer 104 can determine that the second person is likely to be new to the scene since he/she appeared at or near a portal region.
[0072] Classification of a given image region is based on comparing information stored in the background feature matrices for the image region to a threshold. In some embodiments, the CAS computer 104 has a threshold for each background feature. For example, if the value stored the floor matrix for image region {XJ, yj} exceeds the floor threshold, then image region {xj, yj} is classified as a floor. [0073] In other embodiments, classification depends on comparing the value stored for an image region from two or more background feature matrices. In one embodiment, an image region is classified as an obstruction region if the value stored in the obstruction matrix for that image region exceeds the obstruction threshold, but the value of the floor matrix for that image region does not exceed the floor threshold. In that embodiment, an image region is classified as a floor region and not an obstruction region if the value of the floor and obstruction matrices for that image region both exceed their respective thresholds. Thresholds are discussed below in further detail.
[0074] Figure 4 is a flow chart depicting a background classification method 400 according to one embodiment of the invention. To understand the process of maintaining the background feature matrices, Figure 4 will be described in conjunction with Figures 5-8.
[0075] The CAS computer 104 utilizing the background classification method 400 maintains an object list that includes information about objects that are present in each video frame. The object list may contain identifying information for each object in the video frame, including, for example, the location, color, and size of each object as it appears in the current video frame and as it had appeared in previous video frames. Once added to the object list, the object remains in the object list until removed. First, the object list and the background feature matrices are initialized (step 402). The object list is cleared and the data stored for the image regions in the background feature matrices are reset to zero. In the illustrative embodiment, the background feature matrices store counts of occurrences related to each background feature.
[0076] In one embodiment, the CAS computer 104 then analyzes a series of video frames, one frame at a time. For each video frame, the CAS Computer 104 identifies the objects in the video frame and labels each identified object as unprocessed (step 404). The CAS computer 104 identifies objects using one of several methods commonly known in the art (e.g., background subtraction). In step 406, the CAS computer 104 determines whether there are any unprocessed objects in the object list for the video frame. Unless the video frame does not include any objects or all objects have already been analyzed for the video frame, the CAS computer 104 selects one of the unprocessed objects in the object list for analysis (step 408). After selecting an object (step 408) the CAS computer 104 determines whether the selected object is newly appeared in the scene depicted in the video frame (step 410), i.e. was the object not listed in the object list for the preceding video frame.
[0077] Figure 5A is a video frame 500 depicting a scene at time t. Figure 5B is a video frame 500' depicting the same scene at time t+1, i.e. it is the immediately subsequent video frame captured after the video frame 500 of Figure 5 A. Video frames 500 and 500' are overlaid with gridlines depicting boundaries of image regions for illustrative purposes. The video frame 500 includes a door 502 and no objects. The video frame 500' includes the door 502 and an object 504 standing in the door. Since the object 504 was not in video frame 500, it is not listed in the object list for video frame, and therefore the CAS computer 104 considers the object 504 to be newly appeared.
[0078] The determination of a newly appeared object is one occurrence that results in an update of the portal matrix (step 412). Figure 5C is a schematic drawing 500" of the steps the CAS computer 104 takes to update the portal matrix based on analyzing video frames 500 and 500'. Video frame 500" is overlaid with gridlines depicting boundaries of image regions for illustrative purposes. In response to detecting the newly appeared object 504, the CAS computer 104 increments the count stored for each image region in the portal matrix in which the newly appeared object appears. In 500', the newly appeared object appears in image regions {5-6, 4— 7}. As a result, the CAS computer 104 increments the count stored for each image region {5-6, 4—7} in the portal matrix as indicated in the schematic drawing 500". In addition to updating the portal table (step 412) upon detection of a newly appeared object, the CAS computer 104 also adds the object to the object list (step 414).
[0079] The CAS computer 104 also updates the portal matrix in response to determining that an object is newly disappeared. For example, assume that video frames 500 and 500' were captured and analyzed in reverse order. In video frame 500' at time t, the video frame 500' includes an object 504 standing in the door 502. In video frame 500, the immediately subsequent video frame, at time t+1, the video frame 500 no longer includes the object 504. Since no objects are included in video frame 500, in step 406, when the CAS computer 104 determines whether there are any unprocessed objects, the CAS computer 104 does not locate any unprocessed objects, and proceeds to determine if there are any newly disappeared objects (step 416). In doing so, the CAS computer 104 analyzes the object list to determine whether the preceding frame 500' included any objects that the current video frame 500 does not (step 418). Upon detection of the object 504 listed in the object list of the preceding frame, the CAS computer 104 updates the portal matrix (step 420) in the same fashion as described above and as depicted in the schematic drawing 500". The object 504 is then removed from the object list (step 422). In one embodiment, if the CAS computer 104 detects a newly disappeared object, the CAS computer 104 determines if the object was present in or located near at least one image region from the previous frame that is classified as portal (see below). If the newly disappeared object was not in or near a portal image region, the CAS computer 104 keeps the object in the object list for at least one additional frame to take into account the possibility of the CAS computer 104 incorrectly missing the object in the video frame.
[0080] After a newly appeared object is added to the tracking list (step 414) or after a determination that the object being analyzed is not newly added, the CAS computer 104 updates the floor matrix (step 424). To do so, the CAS computer 104 first determines the bottom or lower boundary of the object and then increments the count stored in the floor matrix for one or more image regions based on the location of the bottom of the object.
[0081] Objects have at least one boundary. In one embodiment, boundaries are lines composed of pixels that could be traced around the exterior of the object. In other embodiments, boundaries are boundary regions that surround particular pixels that make up the object. In some embodiments boundaries are also ascribed labels . For example, if one were to draw a real world- vertical axis (i.e., up and down) through an object in a video frame, a portion of the video frame surrounding a pixel of the object that is furthest along the real world- vertical axis in the down direction is a bottom boundary region. In one embodiment, the bottom boundary region is a single pixel. In another embodiments, the bottom boundary region is the image region that includes such a pixel.
[0082] Determining the real world-vertical axis through an object is not trivial. Often it is assumed that the _y-axis in the image (the jy-axis coincides with pixel columns) corresponds to the vertical axis in the real world. However, this assumption is not always valid — for example, if the camera is rotated about its optical axis then the real world-up could be in any direction. In fact, the up direction may not be a single direction at all.
[0083] Under a general perspective imaging model, a set of parallel lines intersects at a common point called the vanishing point. Every set of parallel lines intersects at a different vanishing point. One particular set of parallel lines, the set of all vertical lines, intersects at a point called the vertical vanishing point. The location of vertical vanishing point determines the up direction, and consequently the up direction may vary at different pixels in the image.
[0084] Figures 6A-6C are three video frames 600, 600', and 600", respectively, each demonstrating one of three distinct possible vertical vanishing points. Video frame 600 is an example of a video frame in which the vertical vanishing point 602 conceptually lies off of the screen. Video frame 600' is an example of a video frame in which the vertical vanishing point 602' is at infinity, and video frame 600" is an example of a video frame in which the vertical vanishing 602" point lies within the video frame 600" itself. Video frames 600, 600' and 600" each include several objects 604, 604' and 604". Each object 604, 604' and 604" has a vertical axis 606, 606', and 606" that is an extension of a line that conceptually connects the centroids 608, 608' and 608" of the objects 604, 604', and 6004" with the vertical vanishing point 602, 602', and 602".
[0085] In video frame 600, the vertical axes 606 for all objects 604 are generally the same direction, though it varies somewhat depending upon the video frame-horizontal (the direction of pixel rows) position of the object 604. In video frame 600', the vertical axes 606' of the objects 604' are parallel, i.e. they share the same up direction, corresponding to the direction of pixel columns. In contrast to video frames 600 and 600', the vertical axes 606" of the objects 604" in video frame 600" vary greatly. For objects 604" located towards the middle of the left-hand side of the video frame 600", the up direction is towards the left. For objects 604" located towards the middle of the right hand side of the video frame 600", the up direction is generally to the right. In video frame 600", up is generally away from the center of the video frame 600". An overhead camera, for example, can create video frame 600".
[0086] Figure 7 is a flow chart of an illustrative computerized method of determining the vertical vanishing point 700 for a series of video frames according to one embodiment of the invention. The CAS computer 104 calculates the vertical vanishing point based on the orientation of a plurality of objects in the series of video frames. Figure 7 should be viewed in conjunction with Figure 8 for an understanding of the vertical vanishing point calculation method 700. Figure 8 is one illustrative video frame to be analyzed according to the method 700. In one embodiment, the vanishing point is continuously updated with each video frame; while in other embodiments, the vanishing point is updated periodically.
[0087] For each video frame analyzed, the CAS computer 104 identifies the objects 802, 804 and 806 that are included in the video frame 800 (step 702). The CAS computer 104 then selects an object 802 to analyze (step 704). For each object, the CAS computer 104 carries out the following analytical steps.
[0088] The CAS computer 104 analyzes the selected object 802 to determine if the object 802 appears to be a person (step 706). To do so, the CAS computer 104 computes the major axis 808 and minor axis 810 of the object 802. This computation is a standard computer vision procedure described in introductory textbooks, such as Robot Vision by Berthold Horn, and published by M.I.T. Press in 1986. Intuitively, the major axis 808 points along the "long" dimension of an object 802 and the minor axis 810 points along the "short" dimension. In the case of people, the major axis 808 generally points in the head-to-foot dimension (i.e., the vertical direction) and the minor axis 810 is perpendicular to that.
[0089] If an object does not have long and short dimensions, then the major axis 808 and minor axis 810 are about the same size and point in random directions. Thus, in order to determine if the selected object 802 is "human-like," the CAS computer 104 computes the ratio of the major axis to the minor axis. If this ratio is big enough, for example, in one illustrative embodiment, greater than 1.5, then the object is considered for further processing.
[0090] If the CAS computer 104 determines that the object 802 is human-like, the CAS computer 104 then determines whether the orientation of the object 802 substantially comports with the current vertical vanishing point estimate 816. In one illustrative embodiment, the CAS computer 104 computes a line 812 that should intersect with the vertical vanishing point 816. This line 812 has the slope of the major axis 808 and contains the centroid 814 of the object. The centroid 814 is typically the point at which the object's major axis 808 and minor axis 810 intersect. In one embodiment, the CAS computer 104 then computes the perpendicular distance 817 between this line 812 and the vertical vanishing point 816. If this distance is small enough, for example, 10 pixels, then the object is processed further. In another embodiment, the CAS computer 104 compares a line 818 that passes through both the vertical vanishing point 816 and the centroid 814 of the object 802. If the angle between of the aforementioned lines, 812 and 818, is large, for example if the object 802 is in a prone position or is in the process of falling, the CAS computer 104 does not use selected object 802 as a new data point in updating the vertical vanishing point 816. If the angle between the two lines is less than a small angle (e.g., 0.1 degrees), then the object is processed further. If the CAS computer 104 has selected the object for further processing, the CAS computer 104 incorporates the object into its calculation of a new vertical vanishing point for the video frame (step 710).
[0091] Steps 704, 706, 708, and 710 are repeated for each object in the video frame. The entire method 700 then repeats with the next video frame.
[0092] To calculate the vertical vanishing point, the CAS computer 104 builds a 3X3 vertical vanishing point matrix, updating the matrix with information from the objects selected for further processing. For the purposes of the matrix, a line in the image is represented by three coefficients (A,B,C). If a point (x,y,w) is on the line, then it satisfies the equation Ax + By + Cw = 0. Note that two-dimensional points are represented by three numbers (so-called homogeneous coordinates). In this representation, point (x,y,l) corresponds to pixel (x,y) and point (x,y,0) corresponds to a point at infinity.
[0093] A set of lines is indexed with subscript /': (Ait Bit C,). If the lines intersect at a single point Xvp, yv , Wv ), the vertical vanishing point, then they all satisfy the line equation A vp + Bjyvp + CjWvp = 0, or in matrix format:
Figure imgf000025_0001
[0094] Note that the previous matrix equation grows in size with more data. Solving this equation using standard least squares techniques, we arrive at the following equation:
Figure imgf000025_0002
[0095] Solving the equation for (xvp, yvp, wvp) yields the location of the vertical vanishing point. The well-known solution to this equation is the eigen- vector that corresponds to the smallest eigen-value. This eigen-vector can be computed using standard linear algebra techniques. This equation, the vertical vanishing point equation, has a constant size (3x3) regardless of the amount of data.
[0096] The CAS computer 104 uses the major axes of the objects selected for further processing as the lines used for populating the vertical vanishing point matrix, i.e. the summations, in the vertical vanishing point equation. To update the equation to accommodate new data, the CAS computer 104 simply adds one value to each of the 9 summations in the left-hand side of the equation. The CAS computer 104 derives these values from the line equation for the major axis 808 of the object 802, for example:
Figure imgf000025_0003
mn' = A,+i B,+Y mn' > [0097] In one embodiment, the updating rule favors newer data more than older data. This strategy is useful in the case when the vanishing point is changing, and the newer data is thus more relevant than the older data. In this case, the update rule would look like
Figure imgf000026_0001
m i+l = oAl+ Bl+x + (\ a)ml'2 :
[0098] where α is a weight parameter that varies from 0 to 1. An α value closer to 1 weights
newer data more, while a value closer to 0 weights older data more.
[0099] Based on the vertical vanishing point, the CAS computer 104 determines the bottom or bottom boundary of an object, after which the floor matrix can be updated accordingly (step 424).
[0100] Figure 9 A is a video frame 900 depicting a first object 902 and a second object 904 in a scene. For illustrative purposes, the video frame 900 is overlaid with gridlines indicating the boundaries of image regions. The second object 904 is standing behind a piece of furniture 906. Figure 9B is a schematic drawing 900', including demonstrative image region boundaries, of the steps the CAS computer 104 takes to update the floor and obstruction matrices based on analyzing video frame 900. It is assumed that the vertical vanishing point in the video frame 900 is at infinity and that the vertical axis 901 corresponds to the image region columns, i.e. towards the top of the page is up and towards the bottom of the page is down.
[0101] For purposes of analyzing the first object 902, the bottom boundary region for the first object 902 appears in image region {6, 9}. As a result, the count stored in location {6, 9} of the floor matrix is incremented. The bottom boundary region of the second object 904 appears in image region {8, 8}. Note that the bottom boundary region of the second object 904 does not correspond to the feet (not shown) of the second object 904. It corresponds however, to the lower most, i.e. furthest in the down direction, portion of the second object 904 that is visible to the camera that is taking the video. Correspondingly, the count stored in floor matrix location {8, 8} is also incremented.
[0102] In other embodiments, additional counts for image regions surrounding the bottom of image may also be incremented. For example, in one embodiment, the counts stored in floor matrix locations {5-7, 9-10} are incremented. Incrementing the counts surrounding regions is useful when the CAS computer 104 is using relatively small image regions, thereby populating the matrix faster.
[0103] Fig. 9B also indicates the changes to the counts stored in the obstruction matrix that an illustrative embodiment of the CAS computer 104 would make based on the video frame of Fig. 9A (step 426). In the illustrated embodiment, for each video frame, locations in the obstruction matrix corresponding to an image region in which an object is present, other than the object's bottom boundary region, is incremented. For example, Fig. 9B indicates that the counts stored in obstruction matrix locations {5-6, 5-8} and {5,9} would be incremented in updating the obstruction matrix (step 426) for the first object 902. Similarly, the counts stored in obstruction matrix locations {8, 6-7}, and {9, 7} would be incremented in updating the obstruction matrix (step 426) or the second object 904. In an alternative embodiment, the obstruction matrix location associated with the object's bottom boundary region is also incremented.
[0104] After the CAS computer 104 finishes analyzing an object, the CAS computer 104 determines whether any other objects are present in the video frame that have not been analyzed (step 406). If no additional objects are left to analyze, the video frame determines if there are any newly disappeared objects (step 416) as described above with reference to Figure 5. After all newly disappeared objects are analyzed, the CAS computer 104 advances to the next frame in the series of video frames (step 428), and the process repeats beginning with step 404, by identifying the objects in the new video frame.
[0105] Figure 10 is a flow chart of a method for determining whether to track an object in a video frame based at least in part on the classification of image regions in the video frame 1000. The CAS computer 104 utilizes the data accumulated in the background feature matrices to classify image regions.
[0106] In determining whether to track objects in a video frame, the CAS computer 104 first selects an object in the video frame (step 1002). The CAS computer 104 then determines whether the selected object is on the floor (step 1004), i.e. is the bottom boundary of the object present in an image region that is classified as floor. Determining the bottom boundary of the object is described above with respect to Figures 6-8. The CAS computer 104 then determines whether the image region in which the bottom boundary is classified as floor. In one embodiment, the CAS computer 104 compares the count stored in the image region's location in the floor matrix to a floor threshold. If the count exceeds the threshold, the image region is classified as floor.
[0107] In one embodiment, the floor threshold is determined dynamically based on the following factors: a pre-established estimated value pf, ranging from 0 to 1, representing the proportion of the matrix elements that are being updated correctly; a pre-established estimated value ff, ranging from 0 to 1, representing the fraction of the image regions of the video frame that should be classified as floor regions; the sum Nf of the counts for all image regions in the floor matrix; and the number of image regions Mf in the video frame. In one embodiment, the floor threshold Fthresh is derived from the following equations: PfNf ffMf
(l-Pf )Nf «,..., = (l-ff )Mf
-<*„„, +<*,„ + ln(l -ff ) - ln(ff ) thresh ln(α„,) -ln(α„„, )
[0108] In general, if the floor count for an image region is high enough such that it is more likely that the image region is floor than it is likely that the image region is not floor, then the image region is classified as floor. Each classification has its own threshold such as a floor threshold, an obstruction threshold and a portal threshold. In one embodiment, the thresholds for other classifications are determined by the same equation and substituting p0, f0, M0, and N0 for the obstruction threshold equation and pp, fp, Mp, and Np for the portal threshold equation. In one embodiment the p and f values for the three thresholds are all set initially such that p=0.9 and f=0.5. In another embodiment, the CAS computer 104 updates the f value for each classification based on the number of image regions the CAS computer 104 has classified as belonging to each classification.
[0109] If the lower boundary of the object is present in an image region that the CAS computer 104 classifies as floor, the CAS computer 104 then tests the size of the object to determine if the object is large enough to be tracked by the CAS computer 104 (step 1006). In one embodiment, the size is an absolute number, for example, the object must include at least 25 pixels. In another embodiment, the size might be relative. For example, the CAS computer 104 may compare the length of the major axis of the object with the minor object of the axis as discussed above with respect to determining vertical vanishing points. [0110] If the lower boundary of the object is not on the floor, or if the object is too small to be interesting, the CAS computer 104 determines whether the object is touching an obstruction that might be occluding part of the object (step 1008). For example, referring back to Figure 9, the second object 904 may not be large enough to be tracked because part of the second object 904 is occluded by a piece of furniture. In that case, in the illustrative embodiment, the CAS computer 104 determines whether the lower boundary of the object touches an occluding object. To do so, the CAS computer 104 determines the lower boundary of the second object. The CAS computer 104 then determines the classification of the image region below that in which the lower boundary of the object is present. If that region is classified as an obstruction, that embodiment of the CAS computer 104 tracks the object even if it is not located on the floor or if the object appears to be too small to be tracked.
[0111] If the CAS computer 104 determines not to track the object, the object is discarded (step 1010) and the CAS computer 104 begins analyzing the next object. If the CAS computer 104 determines to track an object, the CAS computer 104 determines whether the object is newly appeared in the scene (step 1012). In one embodiment, the CAS computer 104 maintains an object list that maintains information, including location information, about objects in a scene. Based at least in part on the information stored in the object list, the CAS computer 104 determines whether the object in the video frame being analyzed likely corresponds to an object listed in the object list from the previous video frame. If the object likely corresponds to an object listed in the object list from a previous video frame, the CAS computer 104 continues to track the object (step 1014).
[0112] If the object likely does not correspond to an object in the object list, the CAS computer 104 determines whether the non-corresponding object is present in or located near an image region classified as portal (step 1016). If the non-corresponding object is not present in or located near an image region classified as portal, the CAS computer 104 adds the object to the object list and begins tracking the object (step 1018). The CAS computer 104 then begins analyzing the next object in the video frame (1002), if any remain to be analyzed.
[0113] If the non-corresponding object is present in or located near a portal image region, the CAS computer 104 searches for objects that were present in or located near image regions classified as portals in previous frames and in frames captured by other cameras (1020). For example, if the CAS computer 104 were analyzing two connecting rooms, each with its own camera, and the CAS computer 104 determines that a non-corresponding object is present in an image region classified as portal in the first room, the CAS computer 104 would check the object list to determine if, in a video frame captured by the camera monitoring the second room, an object was present in a region classified as portal in the second room. The CAS computer 104 then determines whether any such objects ("found objects") were found (step 1022). If such there are any found objects, the CAS computer 104 determines that the non-corresponding object corresponds to the found object, and the object continues to be tracked as such (1014). If there is no found object, the CAS computer 104 adds the object to the object list and begins to track it as a newly appeared object (step 1018).
[0114] Referring to Figs. 11 A-14C, a simulation of the background classification system can be prepared by designing a three-dimensional model of a typical surveillance environment with person-sized objects programmed to travel through the model according to random trajectories. The CAS computer 104 then monitors the simulation over a time interval collecting image data as if the graphical output produced in the simulation could be attributed to a plurality of video frames.
[0115] Figs. 11A-C depict three video frames from one such above described simulation of a monitored environment. In Figure 11A, the monitored environment 1100 is devoid of objects. The monitored environment 1100 includes a door 1102, a piece of furniture 1104, and a pillar 1106. In Figures 11B-C, objects 1108 are present (depicted as white rectangular prisms) in the monitored environment. The frame-rate of the simulation is approximately 25 frames per second.
[0116] Figs. 12A-12C, 13A-13C, and 14A-14C are graphical depictions of the state of the background classification matrices populated during the simulation after the CAS computer 104 has analyzed one hundred frames of image data (approximately 5 seconds), one thousand frames of image data (approximately one minute), and ten thousand frames of image data (approximately 10 minutes), respectively.
[0117] Referring to Figs. 12A-C, the shaded region 1200a in Fig. 12A depicts the state of the floor matrix after the analysis of one hundred frames. The shaded region 1200a includes the image regions that have been classified as floor image images. Referring to Fig. 12B, the shaded region 1200b includes the image regions that have been classified as portal image regions. Referring to Fig. 12C, the shaded region 1200c includes the image regions that have been classified as obstruction image regions, but not floor image regions ("non-floor obstruction image regions").
[0118] Referring to Figs. 13A-C, the shaded regions 1300a, 1300b, and 1300c in Figures 13A, 13B, and 13C represent the image regions that are classified as floor regions, portal regions and non-floor obstruction regions after the analysis of one thousand video frames. In comparison to the classifications depicted in Figures 12A-12C, more image regions have been classified in Figs. 13A-C. In Fig. 13 A, most of the floor regions have been identified, including an area on top of the piece of furniture 1104 where objects appear as they move behind the piece of furniture 1104. In Fig. 13B, the image regions around the door 1 102, on either side of the pillar 1106 and around much of the periphery of the environment 1100 have correctly been identified as portal regions.
[0119] Referring to Fig. 14A-C, after ten thousand frames, the classification of image regions, as depicted with the shaded regions 1400a, 1400b and 1400c (floor, portal, and non-floor obstruction regions, respectively), has become more refined, i.e. boundaries of classifications have taken on contoured shapes as opposed to the rougher boundaries noticeable in Figs. 13A- 13C. The classifications depicted in Figs. 14A-C, however, have not significantly changed from their state after the analysis of only one thousand frames, demonstrating that the system can (depending on the number and of objects moving within a monitored environment) generate useful results after only a relatively short monitoring time.
[0120] One skilled in the art will realize the invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. The foregoing embodiments are therefore to be considered in all respects illustrative rather than limiting of the invention. The scope of the invention is not limited to just the foregoing description.
[0121] What is claimed is:

Claims

Claims 1. A computerized method of video analysis, the method comprising: receiving image data for a plurality of video frames depicting a scene that includes at least one of a plurality of background features, wherein (i) each of the video frames comprises a plurality of image regions and (ii) at least one video frame has an object within at least one image region; providing a plurality of background classifications each corresponding to one of the background features in the scene; and assigning one of the background classifications to at least one of the image regions based at least in part on a location of the object relative to the image regions.
2. The method of claim 1 wherein one of the background classifications is a floor.
3. The method of claim 1 wherein the assigning of a background classification to an image region further comprises comparing a value associated with the image region to a floor threshold.
4. The method of claim 1 wherein one of the background classifications is an obstruction.
5. The method claim 1 wherein the assigning of a background classification to an image region further comprises: comparing a value associated with the image region to a floor threshold; and comparing a value associated with the image region to an obstruction threshold.
6. The method of claim 1 wherein one of the background classifications is a portal.
7. The method of claim 1 further comprising: determining for each video frame whether an object has newly appeared in such video frame; and determining the image regions in which the newly appeared objects are present.
8. The method of claim 7 wherein the assigning of a background classification to an image region further comprises counting the number of newly appeared objects that first appeared in the image region.
9. The method of claim 1 further comprising: determining for each video frame whether an object has newly disappeared in such video frame; and determining the image regions in which the newly disappeared objects were last present in a previous video frame.
10. The method of claim 9 wherein the assigning of a background classification to an image region further comprises counting the number of disappeared objects that disappeared from the image region.
11. The method of claim 1 further comprising determining whether to track the object based at least in part on the background classification assigned to at least one of the image regions of the video frame.
12. The method of claim 1 wherein the object further comprises a boundary, the method further comprising the step of determining at least one boundary region that includes the boundary of the object.
13. The method of claim 1 wherein the object further comprises a boundary, the method further comprising the step of determining at least one boundary region that includes at least one of the top, bottom, and side boundaries of the object.
14. The method of claim 13 further comprising determining whether to track the object based at least in part on the image regions in which the at least one boundary region is included relative to the background classification assigned to at least one of (i) such image regions and (ii) another image region in the video frame.
15. The method of claim 1 further comprising determining whether to track the object based at least in part on the size of the object.
16. The method of claim 1 further comprising determining whether to track the object based at least in part on (i) the size of the object and (ii) the image regions in which the object is present relative to the background classification assigned to at least one of (a) such image regions and (b) another image region.
17. The method of claim 1 further comprising: selecting one of the video frames that has an object; and determining whether the object appears in one of the other video frames based at least in part on the background classification assigned to one of the image regions.
18. The method of claim 1 further comprising: selecting one of the video frames that has an object; and determining whether the object appears in one of the other video frames at one of an earlier and later time based on the background classification assigned to one of the image regions.
19. A computerized method of video analysis, the method comprising: receiving image data for a plurality of video frames depicting a scene, wherein at least one video frame has an object within such frame; and determining a vanishing point for such frame based at least in part on one or more characteristics of the object and the vanishing point of other frames.
20. The method of claim 19 further comprising determining the vertical vanishing point for such frame based at least in part on one or more characteristics of the object and the vertical vanishing point of other frames.
21. The method of claim 20 wherein the determining of the vertical vanishing point is further based at least on one of a major axis of the object and a centroid of the object.
22. The method of claim 19, wherein (i) the object further comprises a boundary and (ii) the boundary further comprises plurality of pixels, the method further comprising selecting at least one the pixels of the boundary based at least in part on the vanishing point of the frame.
23. The method of claim 21 wherein (i) the object further comprises a boundary, the boundary having a bottom, and (ii) the boundary further comprises plurality of pixels, the method further comprising selecting at least one pixel that corresponds to the boundary bottom based at least in part on the vertical vanishing point of the frame.
24. A computerized method for video analysis, the method comprising: receiving image data for a plurality of video frames depicting a scene that includes at least one of a plurality of background features, wherein (i) each of the video frames comprises a plurality of image regions, and (ii) at least one video frame has an object within at least one image region, the object having a lower boundary; providing a plurality of background classifications each corresponding to one of the background features in the scene; determining a vertical vanishing point for the at least one frame in which there is an object based at least in part on one or more characteristics of such object and based on the vertical vanishing point of other frames; determining the lower boundary of the object; assigning one of the background classifications to at least one image region based at least in part on the location of the lower boundary of the object in the frame relative to the image region.
25. The method of claim 24 further comprising counting the number of frames in which a lower boundary of an object is present in each image region.
26. The method of claim 25 further comprising counting the number of frames in which an object is present in each image region.
27. The method of claim 26 wherein the assigning of one of the background classifications to at least one image region further comprises comparing the number of frames in which a lower boundary of an object is present in the image region to a floor threshold.
28. The method of claim 27 wherein the assigning of one of the background classifications to at least one image region further comprises comparing the number of frames in which an object is present in the image region to an obstruction threshold.
29. A video analysis system comprising: means for receiving image data for a plurality of video frames depicting a scene that includes at least one of a plurality of background features, wherein (i) each of the video frames comprises a plurality of image regions and (ii) at least one video frame has an object within at least one image region; means for providing a plurality of background classifications each corresponding to one of the background features in the scene; and means for assigning one of the background classifications to at least one of the image regions based at least in part on a location of the object relative to the image regions.
30. A video analysis system comprising: means for receiving image data for a plurality of video frames depicting a scene, wherein at least one video frame has an object within such frame; and means for determining a vanishing point for such frame based at least in part on one or more characteristics of the object and the vanishing point of other frames.
PCT/US2004/029417 2003-09-10 2004-09-09 Method and apparatus for computerized image background analysis WO2005026907A2 (en)

Priority Applications (5)

Application Number Priority Date Filing Date Title
DE602004028136T DE602004028136D1 (en) 2003-09-10 2004-09-09 METHOD AND DEVICE FOR THE COMPUTERIZED PICTURE BACKGROUND ANALYSIS
AU2004272178A AU2004272178B2 (en) 2003-09-10 2004-09-09 Method and apparatus for computerized image background analysis
AT04788655T ATE474293T1 (en) 2003-09-10 2004-09-09 METHOD AND DEVICE FOR COMPUTERIZED IMAGE BACKGROUND ANALYSIS
CA2538294A CA2538294C (en) 2003-09-10 2004-09-09 Method and apparatus for computerized image background analysis
EP04788655A EP1665127B1 (en) 2003-09-10 2004-09-09 Method and apparatus for computerized image background analysis

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US10/659,454 2003-09-10
US10/659,454 US7221775B2 (en) 2002-11-12 2003-09-10 Method and apparatus for computerized image background analysis

Publications (3)

Publication Number Publication Date
WO2005026907A2 WO2005026907A2 (en) 2005-03-24
WO2005026907A3 WO2005026907A3 (en) 2005-05-19
WO2005026907A9 true WO2005026907A9 (en) 2005-06-23

Family

ID=34312699

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2004/029417 WO2005026907A2 (en) 2003-09-10 2004-09-09 Method and apparatus for computerized image background analysis

Country Status (7)

Country Link
US (2) US7221775B2 (en)
EP (1) EP1665127B1 (en)
AT (1) ATE474293T1 (en)
AU (1) AU2004272178B2 (en)
CA (2) CA2538294C (en)
DE (1) DE602004028136D1 (en)
WO (1) WO2005026907A2 (en)

Families Citing this family (54)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6940998B2 (en) * 2000-02-04 2005-09-06 Cernium, Inc. System for automated screening of security cameras
US7221775B2 (en) * 2002-11-12 2007-05-22 Intellivid Corporation Method and apparatus for computerized image background analysis
US7664292B2 (en) * 2003-12-03 2010-02-16 Safehouse International, Inc. Monitoring an output from a camera
US7574416B2 (en) * 2004-01-16 2009-08-11 Saffron Technology, Inc. Distance-based spatial representation and prediction systems, methods and computer program products for associative memories
FR2867864B1 (en) * 2004-03-17 2007-03-02 Automatic Systems METHOD AND INSTALLATION FOR PASSING DETECTION ASSOCIATED WITH A ACCESS DOOR
JP2006079413A (en) * 2004-09-10 2006-03-23 Yokogawa Electric Corp Computer switching apparatus
WO2006115676A2 (en) * 2005-03-30 2006-11-02 Cernium Corporation Video ghost detection by outline
US7801328B2 (en) * 2005-03-31 2010-09-21 Honeywell International Inc. Methods for defining, detecting, analyzing, indexing and retrieving events using video image processing
US7684433B2 (en) * 2005-03-31 2010-03-23 Sony Corporation Method and apparatus for bi-directional communication between analog and digital devices
GB0509225D0 (en) 2005-05-05 2005-06-15 Chroma Therapeutics Ltd Inhibitors of enzymatic activity
GB0509223D0 (en) 2005-05-05 2005-06-15 Chroma Therapeutics Ltd Enzyme inhibitors
US7822224B2 (en) 2005-06-22 2010-10-26 Cernium Corporation Terrain map summary elements
US11361014B2 (en) 2005-10-26 2022-06-14 Cortica Ltd. System and method for completing a user profile
US11403336B2 (en) 2005-10-26 2022-08-02 Cortica Ltd. System and method for removing contextually identical multimedia content elements
US11216498B2 (en) 2005-10-26 2022-01-04 Cortica, Ltd. System and method for generating signatures to three-dimensional multimedia data elements
US20070126876A1 (en) * 2005-11-30 2007-06-07 Eastman Kodak Company Locating digital image planar surfaces
DE102006027120A1 (en) * 2006-06-12 2007-12-13 Robert Bosch Gmbh Image processing method, video surveillance system and computer program
US20080123959A1 (en) * 2006-06-26 2008-05-29 Ratner Edward R Computer-implemented method for automated object recognition and classification in scenes using segment-based object extraction
JP2008035095A (en) * 2006-07-27 2008-02-14 Sony Corp Monitoring apparatus, monitoring system, monitoring method and program
JP4148285B2 (en) * 2006-07-27 2008-09-10 ソニー株式会社 Monitoring device, filter calibration method, and filter calibration program
GB0619753D0 (en) 2006-10-06 2006-11-15 Chroma Therapeutics Ltd Enzyme inhibitors
US20080112593A1 (en) * 2006-11-03 2008-05-15 Ratner Edward R Automated method and apparatus for robust image object recognition and/or classification using multiple temporal views
EP2151128A4 (en) 2007-04-25 2011-11-16 Miovision Technologies Inc Method and system for analyzing multimedia content
US8527655B2 (en) * 2007-05-22 2013-09-03 Vidsys, Inc. Optimal routing of audio, video, and control data through heterogeneous networks
US8264542B2 (en) * 2007-12-31 2012-09-11 Industrial Technology Research Institute Methods and systems for image processing in a multiview video system
WO2009101623A2 (en) * 2008-02-13 2009-08-20 Innovid Inc. Inserting interactive objects into video content
GB2492246B (en) 2008-03-03 2013-04-10 Videoiq Inc Dynamic object classification
US8254629B1 (en) * 2008-08-26 2012-08-28 Adobe Systems Incorporated Methods and apparatus for measuring image stability in a video
US8229228B2 (en) * 2008-09-16 2012-07-24 Robert Bosch Gmbh Image analysis using a pre-calibrated pattern of radiation
GB0903480D0 (en) 2009-02-27 2009-04-08 Chroma Therapeutics Ltd Enzyme Inhibitors
WO2010124062A1 (en) 2009-04-22 2010-10-28 Cernium Corporation System and method for motion detection in a surveillance video
JP5664553B2 (en) * 2009-10-16 2015-02-04 日本電気株式会社 Person clothes feature extraction device, person search device, person clothes feature extraction method, person search method, and program
US9262668B2 (en) * 2010-05-21 2016-02-16 Honeywell International Inc. Distant face recognition system
JP5701005B2 (en) * 2010-10-15 2015-04-15 キヤノン株式会社 Object detection apparatus, object detection method, surveillance camera system, and program
AU2011379626A1 (en) * 2011-10-24 2014-06-12 Helmholtz Zentrum Munchen Deutsches Forschungszentrum Fur Gesundheit Und Umwelt (Gmbh) Method for measuring radiation by means of an electronic terminal having a digital camera
CN102609689B (en) * 2012-02-03 2016-12-28 江苏科海智能系统有限公司 A kind of video track background modeling method based on multiframe statistics
US10289917B1 (en) * 2013-11-12 2019-05-14 Kuna Systems Corporation Sensor to characterize the behavior of a visitor or a notable event
US9053392B2 (en) * 2013-08-28 2015-06-09 Adobe Systems Incorporated Generating a hierarchy of visual pattern classes
US10121254B2 (en) 2013-08-29 2018-11-06 Disney Enterprises, Inc. Methods and systems of detecting object boundaries
US10241988B2 (en) * 2013-12-05 2019-03-26 Lenovo (Singapore) Pte. Ltd. Prioritizing smart tag creation
EP2945373A1 (en) * 2014-05-16 2015-11-18 Misak Ohanesovych Khidirian The security system of objects
US9813605B2 (en) * 2014-10-31 2017-11-07 Lenovo (Singapore) Pte. Ltd. Apparatus, method, and program product for tracking items
US11265317B2 (en) 2015-08-05 2022-03-01 Kyndryl, Inc. Security control for an enterprise network
US11195043B2 (en) 2015-12-15 2021-12-07 Cortica, Ltd. System and method for determining common patterns in multimedia content elements based on key points
US10049462B2 (en) 2016-03-23 2018-08-14 Akcelita, LLC System and method for tracking and annotating multiple objects in a 3D model
US10121515B2 (en) * 2016-06-06 2018-11-06 Avigilon Corporation Method, system and computer program product for interactively identifying same individuals or objects present in video recordings
US9996752B2 (en) * 2016-08-30 2018-06-12 Canon Kabushiki Kaisha Method, system and apparatus for processing an image
US20190347915A1 (en) * 2018-05-11 2019-11-14 Ching-Ming Lai Large-scale Video Monitoring and Recording System
US11126869B2 (en) 2018-10-26 2021-09-21 Cartica Ai Ltd. Tracking after objects
US10452913B1 (en) * 2019-01-31 2019-10-22 Palantir Technologies Inc. Systems and methods for coherent monitoring
US10867178B2 (en) 2019-01-31 2020-12-15 Palantir Technologies Inc. Systems and methods for coherent monitoring
JP7282186B2 (en) 2019-02-12 2023-05-26 コモンウェルス サイエンティフィック アンド インダストリアル リサーチ オーガナイゼーション situational awareness surveillance
US11488290B2 (en) 2019-03-31 2022-11-01 Cortica Ltd. Hybrid representation of a media unit
US20230226998A1 (en) * 2022-01-19 2023-07-20 Nio Technology (Anhui) Co., Ltd. Systems and methods of vehicle surveillance

Family Cites Families (85)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US3740466A (en) * 1970-12-14 1973-06-19 Jackson & Church Electronics C Surveillance system
US4511886A (en) * 1983-06-01 1985-04-16 Micron International, Ltd. Electronic security and surveillance system
GB2183878B (en) * 1985-10-11 1989-09-20 Matsushita Electric Works Ltd Abnormality supervising system
US4819169A (en) * 1986-09-24 1989-04-04 Nissan Motor Company, Limited System and method for calculating movement direction and position of an unmanned vehicle
US5097328A (en) * 1990-10-16 1992-03-17 Boyette Robert B Apparatus and a method for sensing events from a remote location
US5243418A (en) * 1990-11-27 1993-09-07 Kabushiki Kaisha Toshiba Display monitoring system for detecting and tracking an intruder in a monitor area
US5216502A (en) * 1990-12-18 1993-06-01 Barry Katz Surveillance systems for automatically recording transactions
US5305390A (en) * 1991-01-11 1994-04-19 Datatec Industries Inc. Person and object recognition system
WO1992021211A1 (en) * 1991-05-21 1992-11-26 Videotelecom Corp. A multiple medium message recording system
US5237408A (en) * 1991-08-02 1993-08-17 Presearch Incorporated Retrofitting digital video surveillance system
US5164827A (en) * 1991-08-22 1992-11-17 Sensormatic Electronics Corporation Surveillance system with master camera control of slave cameras
JPH0578048A (en) * 1991-09-19 1993-03-30 Hitachi Ltd Detecting device for waiting passenger in elevator hall
US5179441A (en) * 1991-12-18 1993-01-12 The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration Near real-time stereo vision system
US5317394A (en) * 1992-04-30 1994-05-31 Westinghouse Electric Corp. Distributed aperture imaging and tracking system
US5581625A (en) * 1994-01-31 1996-12-03 International Business Machines Corporation Stereo vision system for counting items in a queue
IL113434A0 (en) * 1994-04-25 1995-07-31 Katz Barry Surveillance system and method for asynchronously recording digital data with respect to video data
JPH0811071A (en) 1994-06-29 1996-01-16 Yaskawa Electric Corp Controller for manipulator
CA2155719C (en) 1994-11-22 2005-11-01 Terry Laurence Glatt Video surveillance system with pilot and slave cameras
US6028626A (en) * 1995-01-03 2000-02-22 Arc Incorporated Abnormality detection and surveillance system
US5666157A (en) * 1995-01-03 1997-09-09 Arc Incorporated Abnormality detection and surveillance system
US5729471A (en) * 1995-03-31 1998-03-17 The Regents Of The University Of California Machine dynamic selection of one video camera/image of a scene from multiple video cameras/images of the scene in accordance with a particular perspective on the scene, an object in the scene, or an event in the scene
US5699444A (en) * 1995-03-31 1997-12-16 Synthonics Incorporated Methods and apparatus for using image data to determine camera location and orientation
JP3612360B2 (en) * 1995-04-10 2005-01-19 株式会社大宇エレクトロニクス Motion estimation method using moving object segmentation method
EP0838068B1 (en) * 1995-07-10 2005-10-26 Sarnoff Corporation Method and system for rendering and combining images
WO1997004428A1 (en) 1995-07-20 1997-02-06 Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. Interactive surveillance system
US6002995A (en) * 1995-12-19 1999-12-14 Canon Kabushiki Kaisha Apparatus and method for displaying control information of cameras connected to a network
US6049363A (en) * 1996-02-05 2000-04-11 Texas Instruments Incorporated Object detection method and system for scene change analysis in TV and IR data
US5969755A (en) * 1996-02-05 1999-10-19 Texas Instruments Incorporated Motion based event detection system and method
US5963670A (en) * 1996-02-12 1999-10-05 Massachusetts Institute Of Technology Method and apparatus for classifying and identifying images
US5956081A (en) * 1996-10-23 1999-09-21 Katz; Barry Surveillance system having graphic video integration controller and full motion video switcher
US6526156B1 (en) * 1997-01-10 2003-02-25 Xerox Corporation Apparatus and method for identifying and tracking objects with view-based representations
US5973732A (en) * 1997-02-19 1999-10-26 Guthrie; Thomas C. Object tracking system for monitoring a controlled space
US6456320B2 (en) * 1997-05-27 2002-09-24 Sanyo Electric Co., Ltd. Monitoring system and imaging system
US6295367B1 (en) * 1997-06-19 2001-09-25 Emtera Corporation System and method for tracking movement of objects in a scene using correspondence graphs
US6185314B1 (en) * 1997-06-19 2001-02-06 Ncr Corporation System and method for matching image information to object model information
US6097429A (en) * 1997-08-01 2000-08-01 Esco Electronics Corporation Site control unit for video security system
US6069655A (en) * 1997-08-01 2000-05-30 Wells Fargo Alarm Services, Inc. Advanced video security system
US6188777B1 (en) * 1997-08-01 2001-02-13 Interval Research Corporation Method and apparatus for personnel detection and tracking
US6061088A (en) * 1998-01-20 2000-05-09 Ncr Corporation System and method for multi-resolution background adaptation
US6400830B1 (en) * 1998-02-06 2002-06-04 Compaq Computer Corporation Technique for tracking objects through a series of images
US6400831B2 (en) * 1998-04-02 2002-06-04 Microsoft Corporation Semantic video object segmentation and tracking
US6237647B1 (en) * 1998-04-06 2001-05-29 William Pong Automatic refueling station
AUPP299498A0 (en) * 1998-04-15 1998-05-07 Commonwealth Scientific And Industrial Research Organisation Method of tracking and sensing position of objects
DE69921237T2 (en) 1998-04-30 2006-02-02 Texas Instruments Inc., Dallas Automatic video surveillance system
AUPP340798A0 (en) * 1998-05-07 1998-05-28 Canon Kabushiki Kaisha Automated video interpretation system
JP4157620B2 (en) * 1998-06-19 2008-10-01 株式会社東芝 Moving object detection apparatus and method
US6359647B1 (en) * 1998-08-07 2002-03-19 Philips Electronics North America Corporation Automated camera handoff system for figure tracking in a multiple camera system
US20030025599A1 (en) * 2001-05-11 2003-02-06 Monroe David A. Method and apparatus for collecting, sending, archiving and retrieving motion video and still images and notification of detected events
JP4234863B2 (en) * 1998-12-11 2009-03-04 株式会社アドバンテスト Fail information capturing device, semiconductor memory test device, and semiconductor memory analysis method
US6396535B1 (en) * 1999-02-16 2002-05-28 Mitsubishi Electric Research Laboratories, Inc. Situation awareness system
US6502082B1 (en) * 1999-06-01 2002-12-31 Microsoft Corp Modality fusion for object tracking with training system and method
US6437819B1 (en) * 1999-06-25 2002-08-20 Rohan Christopher Loveland Automated video person tracking system
US6698021B1 (en) * 1999-10-12 2004-02-24 Vigilos, Inc. System and method for remote control of surveillance devices
US6483935B1 (en) * 1999-10-29 2002-11-19 Cognex Corporation System and method for counting parts in multiple fields of view using machine vision
US6549643B1 (en) * 1999-11-30 2003-04-15 Siemens Corporate Research, Inc. System and method for selecting key-frames of video data
CA2391794A1 (en) * 1999-12-06 2001-06-07 Odie Kenneth Carter A system, method, and computer program for managing storage and distribution of money tills
US6574353B1 (en) * 2000-02-08 2003-06-03 University Of Washington Video object tracking using a hierarchy of deformable templates
US6591005B1 (en) * 2000-03-27 2003-07-08 Eastman Kodak Company Method of estimating image format and orientation based upon vanishing point location
US6580821B1 (en) * 2000-03-30 2003-06-17 Nec Corporation Method for computing the location and orientation of an object in three dimensional space
US6850265B1 (en) 2000-04-13 2005-02-01 Koninklijke Philips Electronics N.V. Method and apparatus for tracking moving objects using combined video and audio information in video conferencing and other applications
DE10042935B4 (en) 2000-08-31 2005-07-21 Industrie Technik Ips Gmbh Method for monitoring a predetermined area and system
US6798445B1 (en) * 2000-09-08 2004-09-28 Microsoft Corporation System and method for optically communicating information between a display and a camera
US6813372B2 (en) * 2001-03-30 2004-11-02 Logitech, Inc. Motion and audio detection based webcamming and bandwidth control
US20090231436A1 (en) * 2001-04-19 2009-09-17 Faltesek Anthony E Method and apparatus for tracking with identification
US20030053658A1 (en) * 2001-06-29 2003-03-20 Honeywell International Inc. Surveillance system and methods regarding same
US20030123703A1 (en) * 2001-06-29 2003-07-03 Honeywell International Inc. Method for monitoring a moving object and system regarding same
GB2378339A (en) * 2001-07-31 2003-02-05 Hewlett Packard Co Predictive control of multiple image capture devices.
US7940299B2 (en) * 2001-08-09 2011-05-10 Technest Holdings, Inc. Method and apparatus for an omni-directional video surveillance system
US20030058342A1 (en) 2001-09-27 2003-03-27 Koninklijke Philips Electronics N.V. Optimal multi-camera setup for computer-based visual surveillance
US20030058111A1 (en) * 2001-09-27 2003-03-27 Koninklijke Philips Electronics N.V. Computer vision based elderly care monitoring system
US7110569B2 (en) * 2001-09-27 2006-09-19 Koninklijke Philips Electronics N.V. Video based detection of fall-down and other events
US20030058237A1 (en) * 2001-09-27 2003-03-27 Koninklijke Philips Electronics N.V. Multi-layered background models for improved background-foreground segmentation
AU2002342474A1 (en) * 2001-11-20 2003-06-10 J. Marc Hutchins Facilities management system
US7161615B2 (en) * 2001-11-30 2007-01-09 Pelco System and method for tracking objects and obscuring fields of view under video surveillance
US7123126B2 (en) * 2002-03-26 2006-10-17 Kabushiki Kaisha Toshiba Method of and computer program product for monitoring person's movements
US6847393B2 (en) * 2002-04-19 2005-01-25 Wren Technology Group Method and system for monitoring point of sale exceptions
US7221775B2 (en) * 2002-11-12 2007-05-22 Intellivid Corporation Method and apparatus for computerized image background analysis
WO2004045215A1 (en) * 2002-11-12 2004-05-27 Intellivid Corporation Method and system for tracking and behavioral monitoring of multiple objects moving throuch multiple fields-of-view
WO2004051590A2 (en) * 2002-12-03 2004-06-17 3Rd Millennium Solutions, Ltd. Surveillance system with identification correlation
US6791603B2 (en) * 2002-12-03 2004-09-14 Sensormatic Electronics Corporation Event driven video tracking system
US6998987B2 (en) * 2003-02-26 2006-02-14 Activseye, Inc. Integrated RFID and video tracking system
US20040252197A1 (en) * 2003-05-05 2004-12-16 News Iq Inc. Mobile device management system
US6926202B2 (en) * 2003-07-22 2005-08-09 International Business Machines Corporation System and method of deterring theft of consumers using portable personal shopping solutions in a retail environment
US7049965B2 (en) * 2003-10-02 2006-05-23 General Electric Company Surveillance systems and methods
US20050102183A1 (en) * 2003-11-12 2005-05-12 General Electric Company Monitoring system and method based on information prior to the point of sale

Also Published As

Publication number Publication date
CA2740276A1 (en) 2005-03-24
US20040119848A1 (en) 2004-06-24
WO2005026907A2 (en) 2005-03-24
EP1665127A4 (en) 2008-08-27
AU2004272178A1 (en) 2005-03-24
ATE474293T1 (en) 2010-07-15
US7460685B2 (en) 2008-12-02
EP1665127A2 (en) 2006-06-07
DE602004028136D1 (en) 2010-08-26
WO2005026907A3 (en) 2005-05-19
CA2538294C (en) 2012-11-06
CA2740276C (en) 2014-05-06
US20070211914A1 (en) 2007-09-13
EP1665127B1 (en) 2010-07-14
CA2538294A1 (en) 2005-03-24
AU2004272178B2 (en) 2009-05-07
US7221775B2 (en) 2007-05-22

Similar Documents

Publication Publication Date Title
AU2004272178B2 (en) Method and apparatus for computerized image background analysis
AU2004303397B2 (en) Computerized method and apparatus for determining field-of-view relationships among multiple image sensors
AU2004282902B2 (en) System and method for searching for changes in surveillance video
AU2011201215B2 (en) Intelligent camera selection and object tracking
AU2004282903B2 (en) Method of counting objects in a monitored environment and apparatus for the same

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A2

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BW BY BZ CA CH CN CO CR CU CZ DK DM DZ EC EE EG ES FI GB GD GE GM HR HU ID IL IN IS JP KE KG KP KZ LC LK LR LS LT LU LV MA MD MK MN MW MX MZ NA NI NO NZ PG PH PL PT RO RU SC SD SE SG SK SY TJ TM TN TR TT TZ UA UG US UZ VN YU ZA ZM

AL Designated countries for regional patents

Kind code of ref document: A2

Designated state(s): GM KE LS MW MZ NA SD SZ TZ UG ZM ZW AM AZ BY KG MD RU TJ TM AT BE BG CH CY DE DK EE ES FI FR GB GR HU IE IT MC NL PL PT RO SE SI SK TR BF CF CG CI CM GA GN GQ GW ML MR SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
COP Corrected version of pamphlet

Free format text: PAGES 1/14, DRAWINGS, REPLACED BY NEW PAGES 1/14; DUE TO LATE TRANSMITTAL BY THE RECEIVING OFFICE

WWE Wipo information: entry into national phase

Ref document number: 2538294

Country of ref document: CA

WWE Wipo information: entry into national phase

Ref document number: 2004272178

Country of ref document: AU

WWE Wipo information: entry into national phase

Ref document number: 2004788655

Country of ref document: EP

ENP Entry into the national phase

Ref document number: 2004272178

Country of ref document: AU

Date of ref document: 20040909

Kind code of ref document: A

WWP Wipo information: published in national office

Ref document number: 2004272178

Country of ref document: AU

WWP Wipo information: published in national office

Ref document number: 2004788655

Country of ref document: EP