CN1330275C - Bioassay system based on iris texture analysis - Google Patents

Abstract

The invention relates to a bioassay system based on iris texture analysis, which conquers the defects of existing technic of the absence of iris image quality analysis mechanism support and iris marking image quality evaluation which is not accomplished. The invention is implemented by iris marking image acquisition, automatic iris positioning, evaluation and analysis of iris marking image quality, automatic tracking and detacting of living iris, iris analysis area standardization, polar coordinate transformation, demodulation and expression of iris marking feature, quantizing code generating feature code IrisID, similarity measurement of feature code IrisID, adjudge of similarity measurement. The invention implements basic feature require of idealized system: automatic, non-invasive, friendly human machine operating interface and imaging picture of high quality; no correlation of geometry affine switch and non-linear affine deformation; no dependency of exterior imaging condition; code distributed at random probability; error acceptance rate and error reject rate of high precision.

Description

Bioassay method based on iris texture analysis

[technical field]

The present invention relates to a kind of bioassay method based on iris texture analysis, particularly a kind of realization does not have infringement automatically and obtains high-quality iris image and carry out individual bioassay method based on iris texture analysis, belong to the crossing domain of biostatistics, pattern recognition and computer vision, this method equipment therefor belongs to optical, mechanical and electronic integration, automatic control technology field.

[background technology]

Iris is used for the physiology of individual bioassay and dissects medical features having:

1. iris texture organizes details abundant;

2. it is relevant with the environment in fetal development stage that iris texture is organized the formation of details, has great randomness, breaks up with ovum even be;

3. iris texture tissue signature remains unchanged after back 8 months of birth was to 2 years old;

4. can not change iris texture characteristic with surgical operation at present.

The difficult point of the individual bioassay method of iris research at present and focus are how to realize not have automatically no dependency, image-forming condition no dependence and the high-precision individual bioassay method that infringement and good man machine interface pattern are obtained how much affine transformations such as high-quality iris texture image and translation, convergent-divergent, rotation, pupil elasticity convergent-divergent and non-linear affine deformation.Particularly point out the image-forming condition no dependence and comprise the imaging and focusing quality, imaging noise, image contrast, conditions such as background illumination distribution.Certainly must possess defense mechanism ability to the counterfeit mode.

Mchugh et al " hand-held iris imaging apparatus and method " U.S. Patent number: US6,289,113 B1 " HANDHELD IRIS IMAGING APPARATUS AND METHOD " US PATENT NUMBER:US6, the apparatus and method that 289,113 B1SEP.11.2001 have invented does not have infringement, real-time possesses certain man machine interface feature.It is by imaging len (imaging lens), recessed cold mirror (concave cold mirror), optionally refractive correction mirror (dioptercorrection lens) and near infrared band lighting source, optical video camera composition.This invention also proposes a kind of method that realizes the focus assessment.The non-linear compression method of the energy summation by the analysis image high spectrum is finished the focus assessment.FFT realizes this method for fear of floating-point Fourier conversion, and it has invented following in image space territory wave filter alternative image high frequency spectral domain wave filter real-time implementation.

K(u，v)＝sin(u)sin(v)/π ²UV-sin(2u)sin(2v)/4π ²UV (1)

Bum Chul Kim et al " iris authentication system " U.S. Patent number: US6,594,377 B1 " IRIS RECOGNITIONSYSTEM " US PATENT NUMBER:US6,594,377 B1 Jul, 15,2003 have proposed a kind of do not have the nothing infringement of strict restriction service range, the device of real-time.It has video camera by the control container that can rotate up and down in this container, the ultraviolet cutoff mirror, and cold mirror, several near infrared band lighting sources, distance measurement sensor, control unit is formed.

Bum Chul Kim et al has only described the control relation of a kind of imaging device and this device and control unit, but also lacks a kind of concrete iris image quality control analysis treatment mechanism support when practical application.

In fact, above-mentioned invention mainly all centers on the iris imaging device to be discussed, and the purpose of iris imaging device is to obtain high-quality iris image when practical application, and it must have the combination of image quality analysis treatment mechanism.Realize obtaining the individual bioassay method of reliable iris following requirement must be arranged: analyze and control direct reflection degree, possess defense mechanism ability the counterfeit mode; The analysis and assessment of iris texture picture quality, no dependency, image-forming condition no dependence and the high-precision individual bioassay method of how much affine transformations such as translation, convergent-divergent, rotation, pupil elasticity convergent-divergent and non-linear affine deformation.And require to finish fast.In fact above-mentioned description content is any fundamental requirement of individual bioassay method.

R.P.Wildes et al " iris authentication system and the method for not having infringement automatically " China Patent No.: 95195628.0 disclose a kind of nothing infringement, possess the iris imaging device of certain man machine interface feature.It is mainly by lighting source, Solid imaging video camera, and diffusion filter and polariser are formed.The reduction of fractions to a common denominator of Laplacian Pyramid Wavelet (Laplce's lamination small echo) band is separated and is extracted multiple dimensioned textural characteristics down, adopts the linear statistical data correlation method of Fisher to measure similarity.

The system of R.P.Wildes et al is an experimentation character, can not satisfy the basic feature requirement to idealized system in actual applications fully.

John Daugman is as pioneer and founder's (" based on the individual bioassay system of iris analysis " U.S. Patent number: 5 of iris recognition technology, 291,560 " Biometric Person Identification System Based On IrisAnalysis " US Patent number:5,291,560, May, 1,1994) invention can realize quick and high-precision iris locating method reliably, and its location operator is a following formula:

max|_/_r∮I(x，y)/2πr|ds

(r，x ₀，V ₀)

(2)

Iris analysis geographic norms polar coordinate transform adopts linear boundary affine.

Adopt 2D Gabor small echo bunch texture feature extraction, it has the minimum resolution under Willie-Heisenberg's indeterminacy theory (Weyl-HeisenbergUncertainty Relation), and it has best time-frequency binding analysis precision.

On complex plane, carry out respectively real imaginary part zero passage (zero-crossing) quantization encoding of regional area being produced iris texture characteristic coding IrisCode ^TMThe tolerance of similarity adopts the Hamming distance (Hamming Distance) of coding, and can be realized fast by logic XOR (XOR) computing.

So far, John Daugman is on personal website Http: ∥ www.cl.cam.ac.uk/users/jgd1000/ is detailed The content of iris identification aspect is disclosedComprising detailed contents such as recognition principle and application.

In fact, the result of John Daugman after to wavelet transformation carries out regional area zero passage (zero-crossing) quantization encoding, and the method for logical difference exclusive disjunction Hamming distance measuring similarity is widely known by the people most, and the many researchs of present other based on this.

Similar W.W Boles and B.Boashash " A Human identification Technique usingimages of iris and wavelet transform " IEEE TRANSACTIONS ON SIGNALPROCESSING, VOL.46, APRIL 1998, and the document has also made up the method for wavelet transformation and regional area zero passage quantization encoding.But it is to be based upon on the 1D signal of equal resolution precision in the iris concentric circular.

John Daugman has described the individual bioassay method of iris texture analysis, and its iris locating method possesses fast and high accuracy, but under the situation of pupil and iris region low contrast and noise jamming, the example of Iris Location failure can take place also.

The affine polar coordinate transform of the linear boundary of standardization can not reflect that the biological characteristic that the pupil convergent-divergent changes forms non-linear iris texture compression and stretching and iris texture distribution characteristics performance lack of uniformity.

The real imaginary part zero passage of John Daugman use carrying out on complex plane regional area quantization encoding is the method for a kind of 4 (2bits) state fuzzy quantization.But this coding and method for measuring similarity have certain dependency and limitation to image-forming condition.

Present other many researchs and this field patent all are the methods that is used for Iris Location with general or generalized edge extracting method.This class research method comprises that with various spatial filters be boundary operator such as LOG (Laplacian of Gaussian), the Gaussian anisotropic filter, and Canny boundary operator etc. is transformed to representative in conjunction with circular Hough.

These class methods do not possess practicality, because fast Iris Location is not only indispensable processing procedure, and its locating speed and positioning accuracy all directly influence system reliability and degree of accuracy.This space-like wave filter and circular Hough conversion are used for Iris Location under the application background environment of reality, can not guarantee system reliability and degree of accuracy at all, and simultaneously it can not the assurance system realizes the rapidity requirement of Iris Location.

Prior, the feature coding and the method for measuring similarity of or data statistics relevant based on data all adopted in other many researchs, but these class methods are to the imaging noise, electron gain, picture contrast, image focusing, background illuminations etc. have high susceptibility, that is, and and the image-forming condition high dependency; Also do not possess and do not have relevant coding randomness probability distribution condition, this class coding and measure are considered to not possess reliable and accuracy under analysis high power capacity sample and actual imaging condition.

[summary of the invention]

The objective of the invention is to overcome the defective that lacks the support of iris image quality analysis mechanisms in the prior art, can't finish the iris texture image quality measure, a kind of mode of automatic nothing infringement is provided, obtain high-quality iris image, and automatic analysis and assessment iris image texture quality, realize the assay method of individual bioassay by analyzing the iris texture image.

Prior, realized the fast and reliable Iris Location; Living body iris detects the defense mechanism ability that possesses counterfeit; The self adaptation regional polar coordinate transform that standardizes; Multiple dimensioned based on biological vision cortex model, the wavelet transformation of multidirectional is used for demodulation and expresses the textural characteristics structure; Broad sense binary quanta basic coding method possesses the randomness probability distribution coding that produce not have relies in conjunction with the full probability method for measuring similarity and to the imaging noise, electron gain, picture contrast, image focusing, outside image-forming condition such as background illumination is not had sensitivity.

Another object of the present invention provides the iris imaging device that matches with this assay method, object distance and movement velocity between its energy continuous measurement user and the device; Radiance/the intensity of real time altering light source output; Automatically remove direct reflection in real time; Auto-focus assessment in real time.

The present invention has realized that the basic feature of Perfected process requires: automaticity, nothing infringement, close friend's man machine operation interface and high-quality image; The no dependency of how much affine transformations such as translation, convergent-divergent, rotation, pupil elasticity convergent-divergent and non-linear affine deformation; Outside image-forming condition no dependence; Randomness probability distribution coding; High-precision false acceptance rate (FAR) and false rejection rate (FRR) performance.

The present invention is realized by following scheme:

A kind of bioassay method based on iris texture analysis, it may further comprise the steps:

(1), the iris texture Image Acquisition, comprising:

A), object distance and movement velocity between continuous measurement user and the device;

B), change the radiance/intensity of light source output in real time automatically;

C), remove direct reflection in real time automatically;

D), auto-focus quality evaluation in real time;

(2), automatic Iris Location;

(3), analysis and assessment iris texture picture quality;

(4), living body iris is from motion tracking and detection;

(5), iris analysis geographic norms polar coordinate transform;

(6), iris texture characteristic is expressed in demodulation;

(7), the quantization coding produces feature coding IrisID;

(8), feature coding IrisID measuring similarity.

(9), measuring similarity judgement;

Real-time auto-focus quality evaluation of the present invention may further comprise the steps:

(1) is being the center with the evaluation point, having and define mathematic expectaion, variance or standardization variance in the zone of partial analysis yardstick;

(2) to each local neighborhood pixels analyzed area iterative step (1) in the global analysis image space territory, obtain each partial analysis zone variance or standardization variance, produce all local contiguous analyzed area variances or the set of standardization variance thus;

(3) variance or the set of standardization variance produce focus assessed value or standardization focus assessed value in the accumulation step (2);

(4) the focus assessed value of standard best focus image is gone the assessment focus assessed value that obtains as standard, obtain unified focus quality assessment value, or direct code requirement focus quality assessment value;

(5) the discretization numerical computation method of above-mentioned steps is and optimizes the realization of integer code; The integration that local neighborhood pixels analyzed area is adopted is (1D) line integral of (2D) area branch or x (level) direction, and variance or standardization variance can of equal value be defined as standard deviation or standardizing standard deviation.

Automatically quick Iris Location of the present invention may further comprise the steps:

(1), definition is the center with the evaluation point, has the local assessment area Δ S that radius is analyzed yardstick;

(2), definition is contained in the mathematic expectaion and the variance of the pixel in the Δ S;

(3), by assessment area interested is produced the variance set of all points to be assessed and radius to be assessed set with interval sampling mode iterative step (1)-(2);

(4), variance sets definition to be assessed 2 rank rate of change (acceleration) maximum values to being produced in the step (3);

(5), redefine iterative parameter: with the pairing position of rate of change maximum value, 2 rank that searches in the step (4) is the regional area at center, become new assessment area interested, improve the interval sampling precision of point to be assessed and radius to be assessed, improve the analysis precision that radius is analyzed yardstick;

(6), iterative step (1)-(5), be single pixel until the interval sampling precision;

(7), the discretization numerical computation method of above-mentioned steps is realized fast by optimizing the integer code; And the variance in the above-mentioned steps can of equal value be defined as standard deviation.

Analysis and assessment iris texture picture quality of the present invention adopts and the real-time identical method of auto-focus quality evaluation, and its analyzed area is limited at effective iris region that above-mentioned Iris Location obtains.

Living body iris of the present invention detects and may further comprise the steps:

(1), living body iris detects based on dilator among the ophthalmology and expands the biological characteristic of the pupil convergent-divergent variation of about flesh generation, (that is the radiation of light source of varying strength) regulates the pupil convergent-divergent biology rate of change rule of ophthalmology's principle of the incident luminous flux of control external ambient light automatically based on pupil under controllable lighting condition at random;

(2), the living body iris detection can be by the Photobiology characteristic of the mutual contrast variation of the brightness that produces under the image-forming condition of different arrowbands emission wavelength and various combination thereof based on living body iris biological tissue, under controllable different arrowbands emission wavelength and various combination thereof at random, test vital reaction characteristic judges whether to belong to counterfeit with the mutual contrast variation of the brightness under this different wave length and the various combination image-forming condition thereof;

(3), the living body iris detection also can be by the Photobiology characteristic of the mutual contrast variation of the iris region texture that produces under the image-forming condition of different arrowbands emission wavelength and various combination thereof based on living body iris biological tissue, under controllable different arrowbands emission wavelength and various combination thereof at random, test vital reaction characteristic judges whether to belong to counterfeit with the mutual contrast variation of the texture under this different wave length and the various combination image-forming condition thereof.

Iris analysis geographic norms polar coordinate transform of the present invention may further comprise the steps:

(1), be the 2D regional polar coordinate that standardize according to iris outer edge radius that Iris Location obtained and the coordinate transform of central coordinate of circle definition 2D card Deere;

(2), coordinate system transformation by discretization after, definition angle element sample quantity and increment;

(3), the angle element of definition is finished the 2D regional polar coordinate transform that standardizes according to the discrete sampling mode that is adaptive to different radii in the step (2);

(4), direct reflection and high quadrant eyelid, the eyelashes interference sections zone of the low quadrant of coordinate system eliminating are used to the iris texture characteristic analysis.

Demodulation of the present invention express wavelet transformation that iris texture characteristic adopts can for 2D m rank cardinal B-spline wavelets to be the 2DB-small echo to or multiple dimensioned, multidirectional 2D Morlet small echo, 2D m rank cardinal B-spline wavelets is to being the even number m rank cardinal B-spline wavelets with even symmetry (symmetry), and it is right to have a small echo that odd number m rank cardinal B-spline wavelets of odd symmetry (antisymmetry) forms.And 2D m rank cardinal B-spline wavelets is to satisfying m 〉=3 requirements.

Multiple dimensioned, the 2D Morlet wavelet transformation demodulation of multidirectional, and the direct current/DC of shielding 2D Morlet small echo real part responds output, it makes DC response be output as 0 attribute.The relation of design parameter distributes:

(1) the wavelet scale factor is according to 2 ^j(j=0,1,2 ... N) exponential relationship distributes, and the mid frequency position relation that the mid frequency demodulation factor is determined according to the corresponding wavelet scale factor distributes;

(2) the mid frequency demodulation bandwidth factor and the mid frequency demodulation factor distribute according to the constant proportionate relationship, and this constant guarantees that the band of mid frequency demodulation bandwidth is limited to and has the octave of being less than or equal to.

Quantization coding of the present invention produces feature coding IrisID and may further comprise the steps:

(1), definition is based on the method for iris partial analysis zone broad sense binary quanta basic coding, it is judged by following operator:

$\left\{\begin{array}{ccc}1& \mathrm{when}& f\left(\mathrm{WRe}\right)>g\left(\mathrm{WIm}\right)\\ 0& \mathrm{when}& f\left(\mathrm{WRe}\right)\≤g\left(\mathrm{WIm}\right)\end{array}\right.$

Wherein: f, g are the broad sense functionals, its meet the following conditions (c is positive constant):

$\left\{\begin{array}{cc}f\left(\mathrm{cWRe}\right)>g\left(\mathrm{cWIm}\right)& \mathrm{whenf}\left(\mathrm{WRe}\right)>g\left(\mathrm{WIm}\right);\\ f\left(\mathrm{cWRe}\right)\≤g\left(\mathrm{cWIm}\right)& \mathrm{whenf}\left(\mathrm{WRe}\right)\≤g\left(\mathrm{WIm}\right);\end{array}\right.$

(2), by step (1) to all iris partial analysis area sampling parameters, produce and to possess the measurable iris texture characteristic coding of unified standard IrisID.

Iris texture characteristic of the present invention coding (IrisID) measuring similarity adopts the full probability method for measuring similarity, should adopt on the standard angle direction cycle cyclic shift coding to produce corresponding different iris texture characteristic coding IrisID and need to identify or the registered feature coding of checking carries out repeatedly the postrotational method for measuring similarity of related angle during the full probability measuring similarity.Final measuring similarity result is the postrotational measuring similarity maximum of above-mentioned repeatedly related angle.

Measuring similarity judgement of the present invention is determined the similarity decision value based on data statistic analysis, the individual judgement standard that it is used to determine.Corresponding measuring similarity result provides acknowledge information greater than the decision value system, otherwise provides NACK messages.

The direct reflection of removing automatically in real time of the present invention is that analysis according to video camera imaging image obtains direct reflection degree analyzing result, adjust the position of lighting source and the incident angle between iris, reach change direct reflection degree, until obtaining the interferential iris image of high-quality areflexia.

Of the present invention a kind of based on the used iris optical imaging device of the bioassay method of iris texture analysis, it includes housing, imaging len is arranged in the housing, video camera, and be sequentially arranged on the optical axis, it is characterized in that: be provided with vertical plane incident cold mirror before the optical imaging lens, distance measurement sensor and lighting source are set in the housing, described lighting source is adjusted the positioner of lighting source position and the diffusion filter of sealing lighting source and is formed by visible light-near infrared band lighting source.

Described visible light-near infrared band lighting source further comprises: forms by the surface LED chip array (LED chips) of emission wavelength in the 400-900nm scope and the energy driver that can change the radiation of light source brightness/intensity, and the synthetic and control that can be combined by the independent narrow emission wavelength that decomposes of the emission wavelength of described visible light-near infrared band lighting source.

Above-mentioned visible light-near infrared band lighting source drives the radiance/intensity that is provided as picture iris texture image by energy driver, and the variation of pupil convergent-divergent realizes that living body iris detects under the controllable at random different illumination light intensity conditions.Realize that at independent different arrowbands emission wavelength that decomposes and the Photobiology characteristic variations under the various combination image-forming condition thereof living body iris detects in addition.

Although above-mentioned visible light-near infrared band lighting source emission wavelength is chosen in the 400-900nm scope.But the inventor is verified and the optimal choice of the emission wavelength of affirmation various objectives ground lighting source.

(1) be purpose for carrying out individual bioassay with the graphical analysis of imaging iris texture, optimal lighting source emission wavelength is in the 720-800nm bandwidth range;

(2) realize that for changing with the pupil convergent-divergent it is purpose that living body iris detects, optimal lighting source emission wavelength is in the 400-680nm bandwidth range.

(3) be purpose for realizing that with the Photobiology characteristic variations under the different image-forming conditions living body iris detects, the emission wavelength of optimal lighting source is different arrowbands emission wavelength and various combination thereof that to have typical half-peak value ripple be 30nm in the 400-900nm bandwidth range.

The positioner of the adjustment lighting source position of iris optical imaging device is control motor or choice of location/switch, and light source connection control motor drive shaft or choice of location/switch are to adjust the position of light source.

Distance measurement sensor is when recording different object distances, and the radiance/strength relationship of light source output is I=(D/D _o) ²I _o

Wherein: I _oFor being preset in object distance D _oNormal brightness/the intensity that defines during the place, D and I are object distance and the corresponding radiance/intensity that distance measurement sensor PSD successive dynamic measuring obtains.

The object distance that obtains by Continuous Tracking PSD kinetic measurement, and determine the radiance/intensity of corresponding light source output according to following formula, the adjusting of radiance/intensity is regulated current/voltage output by energy driver and is finished.

Energy driver output regulated value control method presets object distance and corresponding energy driver output regulated value attribute of a relation table for producing, and realizes by the quick look-up table to this attribute list then.

The iris optical imaging device has voice alerting unit and photoelectricity Tip element in the housing.Can make this device easy operating and convenient the use.

The iris optical imaging device, vertical plane incident cold mirror is recessed surperficial cold mirror or plane cold mirror, and be provided with single (or two) eyes vision coverage area guiding sign on the cold mirror surface, the sight line of guiding user realizes that user is with respect to the self-adjusting at imaging viewing field center or self-align.

Iris optical imaging device, video camera are CCD or the CMOS video cameras that has responsive in the 400-1000nm wave-length coverage.

The iris optical imaging device, but optical imaging lens is imaging len or the fixed focal length imaging lens with varifocal focusing function.

Iris optical imaging device, distance measurement sensor are infrared ray or ultrasonic distance measuring element.

Bioassay method based on iris texture analysis of the present invention has following beneficial effect:

1) the present invention has the support of iris image quality control analysis mechanism, to the iris texture image quality measure, obtains high-quality iris image;

2) iris locating method of the present invention can avoid taking place the Iris Location failure that the situation of pupil and iris region low contrast and noise jamming causes, and possesses reliable accuracy fast;

3) living body iris detection method of the present invention possesses the reliable defense mechanism ability to the counterfeit mode;

4) the self adaptation of the present invention regional polar coordinate transform that standardizes, the biological characteristic that reflection pupil convergent-divergent changes forms non-linear iris texture compression and stretching and iris texture distribution characteristics performance lack of uniformity, and the adaptivity of sample mode is guaranteed the degree of accuracy of coordinate system transformation;

5) 2D m of the present invention rank cardinal B-spline wavelets transfer pair be 2D B-wavelet transformation to or multiple dimensioned, multidirectional 2D Morlet small echo is used for the iris texture characteristic demodulation expresses and has the function of biological vision cortex model;

6) broad sense binary quanta basic coding method of the present invention possesses the randomness probability distribution coding that produce not have relies in conjunction with the full probability method for measuring similarity and to the imaging noise, electron gain, picture contrast, image focusing, outside image-forming condition such as background illumination is not had sensitivity;

7) object distance and movement velocity between the present invention's energy continuous measurement user and the device; Dynamically change the radiance/intensity of light source output in real time by energy driver, reach best balanced unified illumination distribution condition;

8) the iris optical imaging device is provided with the eyes vision coverage area and guides sign on the cold mirror surface, the sight line of guiding user, nothing infringement, close friend's man machine operation interface;

9) direct reflection of removing automatically in real time of the present invention is that analysis according to video camera imaging image obtains direct reflection degree analyzing result, adjust the position of lighting source and the incident angle between iris, reach change direct reflection degree, until obtaining the interferential iris image of high-quality areflexia;

10) visible light-near infrared band lighting source emission wavelength verified and that confirm of the present invention selects to have the optimal image-forming condition of formation.

To and contrast accompanying drawing by embodiment below, the present invention will be further described below.

[description of drawings]

Fig. 1 is a method theory diagram of the present invention;

Fig. 2 is the organigram of iris optical imaging device;

Label declaration among the figure: 200-iris optical imaging device; The 100-CCD photographing unit; The 101-cold mirror; The 102-imaging len; The 103-virtual image; The 104-PSD distance measurement sensor; The 105-positioner; 106-diffusion filter; The 107-lighting source; The 108-voice alerting unit; 109-photoelectricity Tip element; 110-guides sign; The 99-eyes.

[specific embodiment]

Embodiment 1, the present invention are used for the structure of the iris optical imaging device 200 of iris capturing and can see from Fig. 2, it includes housing, be aligned in sequence with vertical plane incident cold mirror 101, optical imaging lens 102, CCD photographing unit 100 on the intravital optical axis of shell, PSD distance measurement sensor 104 is set in the housing, visible light-near infrared band lighting source 107, adjust the positioner 105 of lighting source position and diffusion filter 106, voice alerting unit 108 and the photoelectricity Tip element 109 of sealing lighting source.

Described visible light-near infrared band lighting source further comprises: be made up of the surface LED chip array (LED chips) of emission wavelength in the 400-900nm bandwidth range and the energy driver that can change the radiation of light source brightness/intensity.And the emission wavelength of described visible light-near infrared band lighting source is the synthetic and control that can be combined by the independent narrow emission wavelength that decomposes.In fact in general, a kind of arrowband really of emission wavelength of surface LED chip.

Above-mentioned visible light-near infrared band lighting source drives the radiance/intensity that is provided as picture iris texture image by energy driver, and the variation of pupil convergent-divergent realizes that living body iris detects under the controllable at random different illumination light intensity conditions.Realize that at independent different arrowbands emission wavelength that decomposes and the Photobiology characteristic variations under the various combination image-forming condition thereof living body iris detects in addition.

In general, lighting source radiance/intensity is big more under the same terms, and image quality is high more.But need to consider that the eyes radiation safety needs and wavelength and relationship of contrast, CCD is to the factors such as sensitivity of different wave length.Therefore lighting source radiance/intensity is that strictness is restricted.

Although above-mentioned visible light-near infrared band lighting source emission wavelength is chosen in the 400-900nm scope.But the inventor is verified and the optimal choice of the emission wavelength of the following various objectives of affirmation ground lighting source:

(1) be purpose for carrying out individual bioassay with the graphical analysis of imaging iris texture, optimal lighting source emission wavelength is in the 720-800nm bandwidth range;

(2) realize that for changing with the pupil convergent-divergent it is purpose that living body iris detects, optimal lighting source emission wavelength is in the 400-680nm bandwidth range.In fact, so the white light of 400-680nm bandwidth produces soft visual psychology effect under the natural environment to user.The lighting source of 400-680nm bandwidth can be selected white light LEDs.

(3) be purpose for realizing that with the Photobiology characteristic variations under the different image-forming conditions living body iris detects, the emission wavelength of optimal lighting source is different arrowbands emission wavelength and various combination thereof that to have typical half-peak value ripple be 30nm in the 400-900nm bandwidth range;

In fact, described method synthetic by the independent narrow emission wavelength that decomposes and the control that is combined is to realize that above-mentioned visible light-near infrared band lighting source emission wavelength can select the basis of controlling.

In the time of before user is positioned at the vertical plane incident cold mirror 101 of this device, its reflection 700nm and following wave band, user can clearly be seen the visual feedback virtual image 103 that own eyes 99 form, and the above wave band of transmission 700nm is used to the 100 physics imagings of CCD video camera simultaneously.Ideal this vertical plane incident cold mirror is recessed surperficial cold mirror or plane cold mirror, and be provided with single (or two) eyes vision coverage area guiding sign 110 on the cold mirror surface, the sight line of guiding user realizes that user is with respect to the self-adjusting at imaging viewing field center or self-align.Cold mirror possesses the function of removing in addition, and this can not only make the CCD video camera recover the original wavelength sensitivity, can increase as imaging user face-image, practical application functions such as video simultaneously.

Vertical plane incident cold mirror 101 provides a kind of ideal visual feedback mechanism, and by eyes vision coverage area guiding sign 110, the encouragement user of its nature is aimed at the iris of its eyes in X-Y optic axis center simultaneously.

Diffusion filter 106 provides following effect:

A increases the lighting source light field;

The lighting condition that direction such as b. provides evenly to distribute;

C. alleviate the direct reflection degree;

Certainly further can not cause vision to produce stimulation to user.

Positioner 105 can be adjusted the incident angle between the iris of lighting source phase, covers iris image with avoiding direct reflection to produce.

Distance measurement sensor 104 is infrared ray range measurement device or ultrasonic distance measuring element.It is mainly used in object distance between the iris of measuring user and device, and finishes the translational speed of tracking between them.Object distance D and speed of related movement V between the iris of distance measurement sensor 104 measurement user and device, and dynamically change radiance/intensity that light source is exported.For obtaining the lighting condition that image has unified brightness/contrast to distribute when the different object distances, the radiance/intensity of light source output has following relation when different object distances:

I＝(D/D ₀) ²I ₀ (3)

I wherein ₀For being preset in object distance D ₀Normal brightness/the intensity that defines during the place, D and I are object distance and the corresponding radiance/intensity that the PSD successive dynamic measuring obtains.

Therefore should adopt the object distance that obtains by Continuous Tracking PSD kinetic measurement, and determine radiance/intensity that corresponding light source is exported according to following formula, the adjusting of radiance/intensity is regulated current/voltage output by energy driver and is finished.When practical application, can preset object distance and regulate the attribute of a relation table of output valve with corresponding energy driver, realize by quick look-up table (lookup table) then this attribute list.

When the approaching device of user, object distance D and speed of related movement V between the iris of PSD Continuous Tracking measurement user and the device.Be in the normal range of operation when movement velocity V is in static relatively while object distance D, enter next process, otherwise voice suggestion 108 and photoelectricity prompting 109 prompting user are adjusted relative distance and reduced speed of related movement.

When the eyes cornea of user with wear glasses (contact lens or other direct reflection), strong direct reflection can when producing incident, the priming illumination light source often take place.In this case, a large amount of gray scale full scale pixels can appear in image.

System analyzes the minute surface degree of reflection automatically:

$\mathrm{SUM}=\underset{x=1}{\overset{N}{\mathrm{\Σ}}}\underset{y=1}{\overset{N}{\mathrm{\Σ}}}V(x,y)$

V(x，y)＝1；when?I(x，y)＝T

V(x，y)＝0；when?I(x，y)＜T (4)

Wherein: T is a pixel grey scale full scale threshold value; (x y) is image pixel value to I.SUM is the degree of reflection analysis result.When analysis result when setting direct reflection extent control standard, lighting source adjusted by the control motor or selection/switch is adjusted at the lighting source of diverse location with the incident angle between the iris that changes light source and user, change the direct reflection degree by incident angle, until obtaining the interferential iris image of high-quality areflexia.

The invention describes a kind of based on the theoretical real time imaging focus method for evaluating quality that defocuses model that produces of optical diffraction.According to the optical defocus theoretical model:

F(x，y)＝I(x，y)*G _σ(x，y) (5)

G δ (x y) is the Gaussian function (Gaussian) with spatial spread yardstick δ, I (x y) is image function, ^*Expression 2D convolution of functions.This model representation out-of-focus image is that (x, y) (x, y) fuzzy back produces original image I through G δ.

The degree that its spatial spread yardstick δ decision image is defocused, in other words, be exactly with image I (x, y) in each pixel be that the neighborhood pixels zone at center is weighted on average (smoothly) degree.Can calculate corresponding focus level by analyzing δ to the degree that defocuses of image.

In the yardstick T of global analysis, so that (x y) is the center, when Δ T is the partial analysis zone, is appreciated that Δ T is the partial analysis yardstick, and it is used for the sensitivity that reaction pair has the Gaussian optical defocus function of level and smooth yardstick δ.

Certainly the described graphical analysis of the yardstick T of global analysis zone be tool optionally, as whole or local graphical analysis zone, center.According to imaging len optical diffraction theory, this partial analysis zone is for defocusing circle or circle of confusion in theory, that is, this partial analysis zone is that radius is the circle of Δ T.In order to simplify following description, only consider square (certain circular polar coordinate also can be understood by equivalence) situation:

A is when δ → 0+ (forward convergence 0):

$\underset{\mathrm{\σ}\→0+}{\lim }I(x,y)*G_{\mathrm{\σ}}(x,y)=I(x,y)---\left(6\right)$

(6) formula shows when δ → 0+ (forward convergence 0), and (x y) is equivalent to the delta function to G δ, is original image to the optical defocus image in local neighborhood pixels zone.

When b satisfies Δ T/ δ → 0+ as δ:

$\underset{\mathrm{\ΔT}/\mathrm{\σ}\→0+}{\lim }I(x,y)*G_{\mathrm{\σ}}(x,y)=1/4\mathrm{\Δ}{T}^2{\∫}_{x-\mathrm{\ΔT}}^{x+\mathrm{\ΔT}}{\∫}_{y-\mathrm{\ΛT}}^{y+\mathrm{\ΔT}}I(x,y)\mathrm{dxdy}---\left(7\right)$

(7) formula shows the equalization that the optical defocus in local neighborhood pixels zone is modeled as local neighborhood pixels zone when Δ T/ δ → 0+,, from the viewpoint of frequency analysis, only has direct current DC composition that is.

The mean square deviation of wave aberration itself has clear physical meaning, promptly, it is the picture element difference quality index that is based upon on the optical diffraction theoretical basis, the inventor particularly points out and adopts picture element difference quality index can not only reflect the degree of defocusing, the various physical factors that influence picture element difference quality have been expressed in more essential it, in fact, it also reflects as other physical factors such as motion blurs.

Realize real time imaging focus appraisal procedure following steps based on above analysis:

A) in the yardstick T of global analysis, so that (x y) is the center, and Δ T is in the local neighborhood pixels analyzed area, and definition mathematic expectaion E and picture element are poor.Wherein the picture element difference may be defined as following type: standard variance μ or standardizing standard variance μ ^*

Q is an ADC full scale quantization step gradient:

$\left\{\begin{array}{c}{E}_{\mathrm{\ΔT}}\left(I\right(xx,y))=1/4\mathrm{\Δ}{T}^2{\∫}_{x-\mathrm{\ΔT}}^{x+\mathrm{\ΔT}}{\∫}_{y-\mathrm{\ΔT}}^{y+\mathrm{\ΔT}}I(x,y)\mathrm{dxdy}\\ \\ {\mathrm{\μ}}_{\mathrm{\ΔT}}\left(I(x,y)\right)=E_{\mathrm{\ΔT}}{[I(x,y)-E_{\mathrm{\ΔT}}\left(I(x,y)\right)}^2]---(9-1)\\ {{\mathrm{\μ}}^{*}}_{\mathrm{\ΔT}}\left(I(x,y)\right)=\left[{\mathrm{\μ}}_{\mathrm{\ΔT}}\left(I(x,y)\right)\right]/Q^2---(9-2)\end{array}---\left(8\right)\right.$

Understanding above-mentioned theory now intuitively can simply be expressed as with variance and reflect local neighborhood pixels analyzed area relationship of contrast.When δ is very little, variance reflects the textural characteristics contrast of local neighborhood pixels analyzed area itself, and δ when very big with respect to local neighborhood pixels analyzed area scope variance level off to 0, promptly by complete equalization, it is 0 that the result is reflected as local neighborhood pixels analyzed area contrast.

B) a), it is poor to obtain each partial analysis zone mathematic expectaion and picture element, the mathematic expectaion and the picture element difference set that produce all local contiguous analyzed areas thus to each local neighborhood pixels analyzed area iteration in the global analysis image space territory.

C) all local contiguous picture element difference sets that regional ensemble produces of analyzing the b that adds up), and produce focus assessed value Ffocus or standardization focus assessed value F ^*Focus.

D) the focus assessed value Fstandard (being the focus assessed value of standard best focus image) with standard picture removes the assessment focus assessed value Ffocus that obtains as standard, obtains unified focus quality (Focus Quality) by (11a) formula.

Focus?Quality＝(Ffocus/Fstandard)·100％ (10a)

Or direct code requirement focus assessed value F ^*Focus, it has finished focus quality assessment value standardization in essence.

Focus?Quality＝F ^*focus·100％ (10b)

And variance in the above-mentioned steps or standardization variance can of equal value be defined as standard deviation or standardizing standard deviation.

As the method for above-mentioned employing, it reflects the picture element quality evaluation to image in essence.Certainly can understand above-mentioned picture element difference quality index and also can equivalence be defined as the standardizing standard deviation.It is that its nothing depends on picture material itself that the method for code requirement focus assessed value has advantage in essence, can produce to possess unified standardization focus quality assessment value.Simultaneously quality control standard depends on that the partial analysis yardstick changes sensitivity in Δ T to average (or level and smooth) with level and smooth yardstick δ.Therefore this method is the method that preferably adopts.

As above-mentioned description, CCD is also can be by above-mentioned picture element difference quality index reflection to the motion blur of a row interval scan integration guiding, and method is changed into local neighborhood pixels analyzed area (1D) line integral of x (level) direction for local neighborhood pixels analyzed area (2D) area integral in definition (8) formula.

All discrete calculation of said method are optimizes the realization of integer code, can avoid floating-point operation, possesses real-time implementation focus evaluated performance frame by frame, and can be by simple hardware-accelerated realization.

Quick iris locating method described in the invention can avoid taking place the Iris Location failure that the situation of pupil and iris region low contrast and noise jamming causes.It is based on the brightness flop rate in the brightness flop rate of pupil region and the sclera zone Photobiology feature much smaller than the brightness flop rate of iris region.Alternative manner described in the invention can be able to quick realization to precise search (coarse to fine searching) principle by being interpreted as bluring, and it has multiresolution (fuzzy to the accurate) search attribute of simulating human vision.

(a) definition with evaluation point (x y) is the center, and [r-Δ r, r+ Δ r] is the local assessment area Δ S of radius:

$\mathrm{\Δs}={\∫}_{\mathrm{\Φ}1}^{\mathrm{\Φ}2}{\∫}_{r-\mathrm{\Δr}}^{r+\mathrm{\Δr}}\mathrm{rdrd\θ}---\left(11\right)$

I (r, θ) is that (x, polar coordinate representation y), Δ r are that radius is analyzed yardstick to image I;

φ 1, and φ 2 is effective angle scope (counterclockwise);

Wherein parameter is chosen as:

Get φ 1=0 degree when location internal diameter (pupil), φ 2=360 degree, Δ r are got accurate small scale analytical model;

Get two parts zone φ 1=-45 degree, φ 2=45 degree and φ 1=135 degree, φ 2=225 degree, the large scale analytical model that Δ r delivery is stuck with paste respectively at location iris external diameter.

(b) definition is contained in assessment area set mathematic expectaion E and the variance μ of the pixel I (r, θ) in the assessment area Δ S that produces in (a):

$\left\{\begin{array}{c}{E}_{\mathrm{\Δs}}\left(I(r,\mathrm{\θ})\right)=1/\mathrm{\Δs}{\∫}_{\mathrm{\Φ}1}^{\mathrm{\Φ}2}{\∫}_{r-\mathrm{\Δr}}^{r+\mathrm{\Δr}}I(r,\mathrm{\θ})\mathrm{rdrd\θ}---\left(12\right)\\ {\mathrm{\μ}}_{\mathrm{\Δs}}\left(I(r,\mathrm{\θ})\right)=E_{\mathrm{\Δs}}{[I(r,\mathrm{\θ})-E_{\mathrm{\Δs}}\left(I(r,\mathrm{\θ})\right)]}^2---\left(13\right)\end{array}\right.$

Certainly Δ S assessment area set mathematic expectaion E and variance μ also can be defined as in above-mentioned (b):

Comprise with Δ S is the unitary set of assessment area, Δ S unitary set is decomposed into the part set that all contiguous analyzed areas are formed in the Δ S, and producing the mathematic expectaion E and the variance μ of corresponding topical set, add up this part set mathematic expectaion E and variance μ become final Δ S assessment area set mathematic expectaion E and variance μ result.

Certainly can understand above-mentioned variance and also can equivalence be defined as standard deviation.

(c) by to assessment area ROI interested (Region Of Interesting) with interval sampling mode iteration (a)-(b) produce all points to be assessed and radius S set et to be assessed (r, x, assessment area set variance set μ y) (r, x, y).

Defining initialized assessment area ROI interested is all image-regions, and the definition initialization is adopted the ambiguity interval sampling to the ROI assessment area.

(d) (r, x y), define 2 rank rate of change (acceleration) maximum values the variance to be assessed that is produced in (c) to be gathered μ.

R(r，x，y)＝|f(_μ(r，x，y)/_r)|

max(r ₀，x ₀，y ₀)|_R(r，x，y)/_r| (14)

F represents filter function, and it is used to effectively suppress noise jamming.As the Gaussian filter function, medium filtering function etc.; (14) behavioral function of formula have search variance to be assessed set μ (r, x, y) in so that (x0 y0) is the center, and r0 is 2 rank rate of change maximum values of radius.It has utilized the Photobiology feature of the 1 rank brightness flop rate in pupil and sclera zone much smaller than 1 rank brightness flop rate of iris region once more, and the contrast of 1 rank edge brightness rate of change is strengthened.Physically can be regarded as and measure the effective acceleration maximum value.Certainly more high-order rate of change maximum value also can be equal to understanding.

(e) redefine iterative parameter:

(x0 y0) is the regional area at center, becomes new ROI assessment area interested with the pairing position of rate of change maximum value, 2 rank that searches in (d).

Treat evaluation point and radius to be assessed and improve interval sampling precision (promptly lowering the interval sampling pixel quantity).

Improve radius and analyze yardstick Δ r analysis precision.

(f) iteration (a)-(e) step, algorithm end in the interval sampling precision when being single pixel.

Accurately the location should be prior to the iris external diameter for internal diameter in said method (pupil), the contrast that might produce sclera zone and iris because of near infrared band (NIR) image-forming condition down is very low, under the oriented situation of internal diameter (pupil), the central area that can utilize it is as location iris external diameter assessment area ROI interested, to realize very fast and reliable and stable location.

Said method can be accelerated realization by optimizing the integer code, and test is at Pentium 800MHz CPU, and the 640X480 pixel image is consuming time only to be 150ms.Its computation complexity has reduced more than 1000 times in conjunction with circular Hough alternative approach than traditional boundary operator at least.

Certainly can understand, analysis and assessment iris texture picture quality adopts the method for above-mentioned figure image focus quality evaluation automatically, and its analyzed area is limited at effective iris region that above-mentioned Iris Location obtains.

The basic goal of individual bioassay method is definite individual identity of precision, possesses the defense mechanism ability to the counterfeit mode.It is the major issue of individual bioassay method reliability itself that living body iris detects.The counterfeit detection should follow the ophthalmology and the Photobiology principle is carried out.

Living body iris detects based on dilator among the ophthalmology (dilator muscle) and expands the biological characteristic of the pupil convergent-divergent variation of about flesh (sphincter muscle) generation, the physical size that shows as pupil diameter constantly changes with corresponding iris outward flange diameter proportion, even show as iridodonesis (hippus) under the priming illumination condition.And under controllable lighting condition at random, the pupil convergent-divergent changes corresponding biometrics rule.Described controllable different lighting condition (that is the radiation of light source of varying strength) at random down the biological rate of change rule of pupil convergent-divergent be ophthalmology's principle of regulating the incident luminous flux of control external ambient light based on pupil automatically.The light source that should be appreciated that above-mentioned controllable different radiant intensity at random is to be lighting source in the 400-680nm bandwidth range for emission wavelength.When

The ratio ρ of the outer edge diameter of definition iris:

ρ＝rp/ri

By following the tracks of ρ rate of change rule under different lighting conditions, test the vital reaction characteristic with this.Principle usually, relatively under the intense radiation lighting condition, it is obvious to expand the pupil contraction effect that about flesh produces, and corresponding ratio ρ tends to minimizing, on the contrary increase.In fact, above-mentioned visible light wave range lighting source is driven by energy driver provides controllable at random illumination light intensity to detect to realize living body iris.Above-mentioned test vital reaction characteristic must the condition of meeting the demands be Iris Location speed.Pupil convergent-divergent upper limiting frequency is in general added up about 4Hz, i.e. 250ms.Iris Location speed of the present invention can meet the demands.

Method further, living body iris detect can by based on living body iris biological tissue in the image-forming condition of different arrowbands emission wavelength (typical half-peak value ripple is 30nm) and various combination thereof the Photobiology characteristic of the mutual contrast variation of brightness of generation down.Under controllable different arrowbands emission wavelength and various combination thereof at random, test vital reaction characteristic judging whether to belong to counterfeit with the mutual contrast variation of the brightness this different wave length and various combination image-forming condition thereof under.Should be appreciated that the above-mentioned different narrow emission wavelength and the emission wavelength of various combination thereof are in the 400-900nm bandwidth range.In fact, living body iris detects based on following Photobiology characteristic:

(1) under the image-forming condition of different arrowbands emission wavelength and various combination thereof, be that main effects causes sclera (sclera) zone to produce the Photobiology characteristic of corresponding brightness Bsclera changes in contrast because of oxygenated haemoglobin (oxygenatedhemoglobin) has different absorbances.

(2) under the image-forming condition of different arrowbands emission wavelength and various combination thereof, be the Photobiology characteristic that main effects produces corresponding brightness Bskin changes in contrast because of there is different reflectance in top layer skin (skin) zone.

(3) under the image-forming condition of different arrowbands emission wavelength and various combination thereof,, iris region is combined as the Photobiology characteristic that main effects produces corresponding brightness Biris changes in contrast because of having different absorption/scatterings/reflectance.

Certainly in the practical application, should consider the sensitivity of CCD to different wave length and various combination thereof, light radiation intensity and automatic gain control (AGC) etc. influence the factor of imaging, but in view of above-mentioned influence factor to sclera regional luminance Bsclera, top layer skin area brightness Bskin, reaching iris region brightness Biris is to possess equivalence in identical narrow emission wavelength or its like combinations imaging image, i.e. no dependence.{ Bsclera, Bskin, Biris} have formed for the mutual relationship of contrast of the brightness of identical narrow emission wavelength or its like combinations, are appreciated that the mutual relationship of contrast of this brightness do not have the above-mentioned imaging factor that influences that depends in brightness.In fact, living body iris detects and just is being based under the image-forming condition of controllable at random different arrowbands emission wavelength and various combination thereof, the mutual relationship of contrast of brightness is changed judging whether to be counterfeit.

Based on same principle, living body iris detects also can be by the Photobiology characteristic of the mutual contrast variation of the iris region texture that produces under the image-forming condition of different arrowbands emission wavelength and various combination thereof based on living body iris biological tissue in addition.Under controllable different arrowbands emission wavelength and various combination thereof at random, test vital reaction characteristic judging whether to belong to counterfeit with the mutual contrast variation of the texture this different wave length and various combination image-forming condition thereof under.The mutual contrast of texture can adopt above-mentioned iris texture image quality measure result comprehensively to contrast.This method is to be based upon on the basis of iris texture image picture element difference, and the iris texture image imaging picture element difference under the identical image-forming condition is that living body iris biological tissue embodies the comprehensive contrast effect that the absorption/scattering/reflectance with wavelength dependency is combined to form.

Use may command principle at random in the lay special stress on said method, under condition like this, provide exponential defense mechanism ability.Carry out 4 kinds of different arrowbands emission wavelengths and various combination thereof as employing, the combination vital reaction characteristic test of 5 STOCHASTIC CONTROL principles, it judges that to counterfeit reliability has improved 4 ⁵=1024 times!

Iris outer edge radius and the central coordinate of circle that above-mentioned Iris Location obtains be respectively (rp, xp, yp) and (ri,, xi, yi).Iris original image I (x, y) institute's employing coordinate system is 2D card Deere coordinate (2D Cartesian coordination), by the regional polar coordinate that standardize (normal Polar coordination) be transformed to P (X (and r, θ), Y (r, θ)).

I(x，y)：P(X(r，θ)，Y(r，θ)) (15)

Wherein: r, θ is the element of polar radius of standardization and angle.

They are turned to r ∈ [0,1] by standard, θ ∈ [0,2 π].

X(r，θ)＝x _p(θ)+f(r)[x _i(θ)-x _p(θ)]；

Y(r，θ)＝Y _p(θ)+f(r)[y _i(θ)-y _p(θ)]；

x _p(θ)＝x _p+r _pcos(θ)

y _p(θ)＝y _p+r _psin(θ)

x _i(θ)＝x _i+r _icos(θ)

y _i(θ)＝y _i+r _isin(θ)

F (r)=∑ a _nr ⁿN=0,1...N are the finite element integer; a _nBe corresponding fitting coefficient;

F (r) is the iris radius deformation function that is formed by the biological characteristic that the irregular pupil convergent-divergent that dilator and the about flesh of expansion produce changes, it causes the iris texture compression and stretches is non-linear affine deformation, on the other hand, the iris texture distribution characteristics also shows as lack of uniformity.Be appreciated that it obtains this function f (r) by statistics iris radius deformation and distribution characteristics match;

After coordinate system transformation was by discretization, sample quantity M and the increment Delta θ (Rad) of definition angle element θ were respectively:

M=[2 π]; [] expression rounds; Δ θ=2 π /=1/r;

R is defined as the different radii in effective iris region.

The inventor particularly points out owing to existing in the actual environment application:

1) because of different image-forming conditions, the difference of effective iris region resolution in the image;

2) the element discretization post-sampling density of polar radius of standardization and angle is different, that is, and and in the corresponding thereupon change of sampling density of different radii with angle element sampling density;

Therefore by the above-mentioned coordinate system transform definition, to finish coordinate transform in the discrete sampling mode that the angle element is adaptive to different radii, this method can not only realize adapting to different iris image resolution, and guarantee the degree of accuracy of coordinate system transformation because of the adaptivity of its sample mode, it also hints increases the degree of accuracy that aftermentioned is analyzed.

The above-mentioned coordinate system conversion realizes optics is become doubly, object distance, degrees of offset, the no dependency of how much affine transformations such as pupil elasticity convergent-divergent and non-linear affine deformation.Be summed up as the direct reflection and the interference such as high quadrant eyelid, eyelashes of low quadrant, system gets rid of this part zone and is used to the textural characteristics analysis.

The inventor has reason to think that bidimensional (2D) image is more reasonable than one dimension (1D) Flame Image Process, because all kinds of actual application environment problems such as above-mentioned how much affine transformations and non-linear affine deformation and calculating quantization error, bidimensional 2D image more can guarantee identical iris texture and be determined by consistent under actual application environment.The 1D image is then uncertain, and it depends on above-mentioned actual application environment strongly.

" image texture characteristic " can be understood as it in the indeterminacy minimum in conjunction with under the resolution analysis window precision theory, express in conjunction with the embedded modulation (modulation) in local space territory by spatial frequency domain.Be defined as demodulation (demodulation) expression according to above-mentioned understanding to forming " textural characteristics " inverting or inverse transformation.

Pattern recognition most important and basic demand in application in practice is that the measuring similarity of feature coding that the image texture characteristic demodulation is expressed and feature coding is to the imaging noise, electron gain, picture contrast, image focusing, external environment image-forming conditions such as background illumination are not had sensitivity, that is image-forming condition no dependence.Can possess the no cross correlation of generation (independence mutually) probability distribution coding on the other hand.

The present invention's application is multiple dimensioned based on biological vision cortex model, and the wavelet transformation of multidirectional is used for demodulation (demodulation) and expresses the textural characteristics structure, and wavelet transformation is used to finish in conjunction with time-frequency/spatial frequency domain multiresolution band reduction of fractions to a common denominator analyses.Existing at present extensive studies achievement of wavelet transformation and practical application, as LOG, Haar, Gabor small echo etc.

The wavelet transformation that the present invention adopts can be 2D m rank cardinal B-spline wavelets (Cardinal B-spline wavelet), i.e. 2DB-small echo or 2D Morlet small echo.

M (positive integer) rank cardinal B-batten Nm small echo ψ m has the most effective computational complexity and effectiveness to or hardware-accelerated realization discrete with software, because it is to have minimum tight support set (compactly supported).Support set for m rank cardinal B-batten Nm and small echo ψ m is:

$\left\{\begin{array}{c}\mathrm{suppNm}=[0,m];\\ \mathrm{supp\ψm}=[\mathrm{0,2}m-1];\end{array}\right.$

More useful characteristic is that even number m rank cardinal B-spline wavelets ψ m has even symmetry (symmetry), and odd number m rank cardinal B-spline wavelets ψ m has odd symmetry (antisymmetry).That is,

And high m (typical m 〉=) rank cardinal B-spline wavelets ψ m extremely approaches the time-frequency/spatial frequency domain minimum of indeterminacy theory in conjunction with the resolution precision, promptly, the high m of 2D rank cardinal B-spline wavelets ψ m approaches following 2D Morlet small echo precision, and the high even number m of 2D rank cardinal B-spline wavelets ψ m approaches M _Re, the high odd number m of 2D rank cardinal B-spline wavelets ψ m approaches M _Im

The generating function or the basic small echo (basic wavelet) of definition 2D Morlet wavelet transformation are:

$\left\{\begin{array}{c}{M}_{\mathrm{Re}}(x,y)=\cos \left(2\mathrm{\π\ωx}\right)\exp [{-(x}^2/2\mathrm{\σ}{x}^2+y^2/2\mathrm{\σ}{y}^2)]\\ M_{\mathrm{Im}}(x,y)=\sin \left(2\mathrm{\π\ωx}\right)\exp [-(x^2/2\mathrm{\σ}{x}^2+y^2/2\mathrm{\σ}{y}^2)]\end{array}---\left(16\right)\right.$

Through rotation, translation becomes small echo bunch set behind the stretching:

$\left\{\begin{array}{c}{M}_{\mathrm{Re}}(a,x^{\′},y^{\′},\mathrm{\θ})=\cos (2\mathrm{\π\ω}{x}^{\′}/a)\exp \left[-{(x^{\′}/a)}^2/2\mathrm{\σ}{x}^2+{(y^{\′}/a)}^2/2\mathrm{\σ}{y}^2)\right]\\ M_{\mathrm{Im}}(a,x^{\′},y^{\′},\mathrm{\θ})=\sin (2\mathrm{\π\ω}{x}^{\′}/a)\exp [-({(x^{\′}/a)}^2/2\mathrm{\σ}{x}^2+){(y^{\′}/a)}^2/2\mathrm{\σ}{y}^2\left)\right]\\ \\ x^{\′}=(x-x0)\cos \mathrm{\θ}+(y-y0)\sin \mathrm{\θ}\\ y^{\′}=-(x-x0)\sin \mathrm{\θ}+(y-y0)\cos \mathrm{\θ}\end{array}---\left(17\right)\right.$

Wherein: a is the wavelet scale factor, and x0, y0 are shift factor, and θ is the angle twiddle factor, and σ x, σ y are mid frequency demodulation bandwidth factor, and ω is the mid frequency demodulation factor.

Small echo bunch has been formed and have been possessed self-enclosed property and completeness, but it does not have orthogonality, as biological vision cortex model exterior syndrome it be fully redundant.According to the requirement based on biological vision cortex (visual cortex) model, the relation of design parameter is distributed as:

(2) wavelet scale factor a is according to a=2 ^j(j=0,1,2 ... N) exponential relationship distributes, and (ω/a) definite mid frequency position relation distributes mid frequency demodulation factor ω according to corresponding;

(2) mid frequency demodulation bandwidth factor σ x, σ y and mid frequency demodulation factor ω distribute according to the constant k proportionate relationship, and this constant k guarantees that the band limit (Band-Limited) of mid frequency demodulation bandwidth is for having the octave of being less than or equal to.

The inventor particularly points out owing to 2D Morlet small echo real part M _RePossess even symmetry, depend on image-forming condition, force shielding small echo real part M for follow-up IrisID (iris texture characteristic coding) is not had _ReDirect current (DC) response output, it can be equivalent to and make DC response be output as 0 attribute.And because of in 2D Morlet small echo imaginary part M _ImEssence possesses odd symmetry, so need not special processing.

Definition iris analysis zone polar coordinate image P (X (r, θ), the equivalent representation P of Y (r, θ)) (x, demodulation y) is expressed as:

$\left\{\begin{array}{c}{W}_{\mathrm{Re}}(a,x0,y0,\mathrm{\θ})=P(x,y)*M_{\mathrm{Re}}(a,x,y,\mathrm{\θ})\\ W_{\mathrm{Im}}(a,x0,y0,\mathrm{\θ})=P(x,y)*M_{\mathrm{Im}}(a,x,y,\mathrm{\θ})\end{array}---\left(18\right)\right.$

Wherein ^*Expression 2D convolution of functions.

In fact, the iris texture characteristic demodulation expression of all regional areas is determined in the complete pointwise of said method energy.

Above-mentioned definition demodulation is expressed in when handling low frequency signal, and a scale factor increases, and mid frequency reduces, and the spatial analysis window is widened automatically; Vice versa, and when handling high-frequency signal, a scale factor reduces, and mid frequency raises, and the spatial analysis window dwindles automatically; Thereby it possesses the signal space cycle that can guarantee comprising equivalence in effective analyst coverage (yardstick) all the time.

The inventor particularly points out to convenient and describes, and following WRe is that (y θ) is the basis for a, x with identical parameter with WIm.To with identical parameter (a, x, y, θ) demodulation that produces is expressed advantage that WRe and WIm carry out corresponding encoded and method for measuring similarity as described below.

The present invention definition has not having the method based on regional area binary quanta basic coding of sensitivity under the above-mentioned external environment image-forming condition, and it is judged by following broad sense binary quanta basic coding operator:

$\left\{\begin{array}{ccc}1& \mathrm{when}& f\left(\mathrm{WRe}\right)>g\left(\mathrm{WIm}\right)\\ 0& \mathrm{when}& f\left(\mathrm{WRe}\right)\≤g\left(\mathrm{WIm}\right)\end{array}---\left(19\right)\right.$

Wherein: f, g are the broad sense functionals, its meet the following conditions (c is positive constant):

$\left\{\begin{array}{cc}f\left(\mathrm{cWRe}\right)>g\left(\mathrm{cWIm}\right)& \mathrm{whenf}\left(\mathrm{WRe}\right)>g\left(\mathrm{WIm}\right);\\ f\left(\mathrm{cWRe}\right)\≤g\left(\mathrm{cWIm}\right)& \mathrm{whenf}\left(\mathrm{WRe}\right)\≤g\left(\mathrm{WIm}\right);\end{array}---\left(20\right)\right.$

In fact, consider in the above-mentioned definition (19) identical parameter (a, x, y, θ), that is, and identical initial phase, no phase contrast.Any textural characteristics signal, the textural characteristics of demodulation method definition (18) corresponding generation is expressed WRe and WIm, and can further prove encoded attributes does not have cross correlation (independence mutually).Will of course be appreciated that described encoded attributes is meant with respect to above-mentioned broad sense binary quanta basic coding method.

R(WRe，WIm)＝<WRe，WIm>＝0；

Wherein:＜presentation code attribute cross-correlation.

Above-mentioned proof procedure is based on M _ReWith M _ImThe essence of no cross correlation is true.Biological vision cortex model itself may also be to have design principle like this with its mutual independence of exterior syndrome.

Can derive above-mentioned broad sense binary quanta basic coding operator generation at last local regional texture feature signal is not had cross correlation (independence mutually) probability distribution coding.

Simultaneously as foregoing description, WRe possesses identical parameter (a, x, y with WIm, θ), can prove also that to local regional texture feature signal the generation of broad sense binary quanta basic coding operator has above-mentioned imaging noise, electron gain, picture contrast, image focusing, external environment image-forming conditions such as background illumination are not had sensitivity, that is, and and the essential advantage of image-forming condition no dependence.

Above-mentioned broad sense binary quanta basic coding operator has to external world the environment image-forming condition and does not have sensitivity, uses so also be applicable to general pattern recognition.

(y θ), produces and to possess the measurable iris texture characteristic coding of unified standard IrisID for a, x to all iris partial analysis area sampling parameters by above-mentioned broad sense binary quanta basic coding method.

Broad sense binary quanta basic coding method of the present invention produces independently any period circulation adjacent states of binary digit primitive encoding position all only has 1 bits of coded to change, and it can avoid adjacent states to produce the multidigit bits of coded by randomly changing effectively.

Pass through foregoing description, be appreciated that broad sense binary quanta basic coding method of the present invention possesses the randomness probability distribution coding that produce not have relies on and to the imaging noise, electron gain, picture contrast, image focusing, external environment image-forming conditions such as background illumination etc. are not had sensitivity, that is, and and the image-forming condition no dependence.

Iris texture characteristic coding (IrisID) measuring similarity of the present invention adopts the full probability method for measuring similarity.This method is from the tolerance probability based on similarity between the theoretical derivation of full probability iris texture characteristic coding.

Full probability measuring similarity between definition iris texture characteristic coding IrisID1 and IrisID2 is:

$\mathrm{Metrlc}(\mathrm{IrisID}1,\mathrm{IrisID}2)=P\left(A\right)=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{p}_i{h}_i=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}P\left(A\right|\mathrm{Bi})P\left(\mathrm{Bi}\right)$

$\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}P1=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}P\left(A\right|\mathrm{Bi})=1---\left(21\right)$

Wherein: N be between iris texture characteristic coding (IrisID) all feature coding bits to quantity;

P (A) is expressed as the full probability with measuring similarity between the right iris texture characteristic coding (IrisID) of N bit.

P (Bi) is expressed as between iris texture characteristic coding IrisID the Bi bit to the independent matching probability of measuring similarity.Be summed up as the present invention and adopt the binary digit primitive encoding, so it only uses independent matching probability on 1 or 0 its mathematical theory of expression.P (Bi) is defined as:

P(Bi)

1 when Bi?IrisID1＝Bi?IrisID2

0 when Bi?IrisID1≠Bi?IrisID2

P (A|Bi) is expressed as between iris texture characteristic coding IrisID the Bi bit to respect to having the right normalization statistical conditional probability that binary encoding possessed of N bit.It is mainly considered:

A. iris texture has great radial dependency, and P (A|Bi) is because of its difference.

B.P (A|Bi) is because of iris texture distributed areas and frequency characteristic difference.

Image because of factor imagings such as low new line of user and device physical positions certainly exists angle rotation problem in actual applications.The symmetrical polar coordinate characteristic of circle that possesses according to iris itself, promptly, it has angle circle symmetric periodic cycle characteristics, after iris texture characteristic demodulation in the polar coordinate image of above-mentioned equivalence was expressed, cycle cyclic shift coding produced corresponding different iris texture characteristics coding IrisID on the standard angle direction.Be readily appreciated that the iris texture characteristic coding IrisID that above-mentioned loop cycle displacement coding produces can equivalence be image angle rotary coding.

Therefore when the full probability measuring similarity, should adopt on above-mentioned standard angle direction cycle cyclic shift coding to produce corresponding different iris texture characteristic coding IrisID and carry out repeatedly the postrotational method for measuring similarity of related angle with registered (Enroll) feature coding that needs to identify (identify) or checking (verify).Final measuring similarity result is the postrotational measuring similarity maximum of above-mentioned repeatedly related angle.

In fact, for be based upon the complete definite characteristic solution mileometer adjustment of the pointwise of having described reach with basis of coding on, this method realizes iris image rotation completely no correlation.

The inventor thinks above-mentioned broad sense binary quanta basic coding method in conjunction with the full probability method for measuring similarity, possesses coding and the method for measuring similarity of distinguishing or data statistics relevant based on data in essence.They possess the randomness probability distribution coding that produce not have relies on and to the imaging noise, electron gain, and picture contrast, image focusing, outside image-forming condition such as background illumination is not had sensitivity.

By determining the similarity decision value based on data statistic analysis, the individual judgement standard that it is used to determine.The data statistic analysis model can adopt Gaussian probability-distribution function f (x), that is, and and f (P (A)).

f(x)＝exp[-(x-x ₀) ²/2σ ²] (22)

Wherein: x=P (A); x ₀Be the statistical mathematics expectation; σ is a statistical variance; P (A) is the measuring similarity full probability.

Determine similarity decision value x, that is, the do not match distribution relation f (x) of rate (FNMR) performance of P (A) and erroneous matching rate (FMR) and mistake, corresponding similarity provides acknowledge information greater than the decision value system, otherwise provides NACK messages.The result is fed back to user by voice or photoelectricity prompting.

By reading above embodiment content, this professional field personnel understand man machine operation interface and the high-quality image that the present invention possesses automaticity, nothing infringement, close friend; The no dependency of geometry affine transformation and non-linear affine deformation; Outside image-forming condition no dependence; Randomness probability distribution coding; High-precision false acceptance rate (FAR) and false rejection rate (FRR) performance.And content that can describe these or close understanding is carried out concrete case enforcement.

Claims (19)

1, a kind of bioassay method based on iris texture analysis is characterized in that may further comprise the steps:

(1), the iris texture Image Acquisition, comprising:

A), object distance and movement velocity between continuous measurement user and the device;

B), change the radiance/intensity of light source output in real time automatically;

C), remove direct reflection in real time automatically;

D), auto-focus quality evaluation in real time;

(2), automatic Iris Location;

(3), analysis and assessment iris texture picture quality;

(4), living body iris is from motion tracking and detection;

(5), iris analysis geographic norms polar coordinate transform;

(6), iris texture characteristic is expressed in demodulation;

(7), the quantization coding produces feature coding IrisID;

(8), feature coding IrisID measuring similarity;

(9), measuring similarity judgement.

2, the bioassay method based on iris texture analysis according to claim 1 is characterized in that: described real-time auto-focus quality evaluation may further comprise the steps:

(1) is being the center with the evaluation point, having and define mathematic expectaion, variance or standardization variance in the zone of partial analysis yardstick;

(2) to each local neighborhood pixels analyzed area iterative step (1) in the global analysis image space territory, obtain each partial analysis zone variance or standardization variance, produce all local contiguous analyzed area variances or the set of standardization variance thus;

(3) variance or the set of standardization variance produce focus assessed value or standardization focus assessed value in the accumulation step (2);

(4) the focus assessed value of standard best focus image is gone the assessment focus assessed value that obtains as standard, obtain unified focus quality assessment value, or direct code requirement focus quality assessment value;

The integration that local neighborhood pixels analyzed area is adopted is (1D) line integral of (2D) area branch or x (level) direction, and variance or standardization variance can of equal value be defined as standard deviation or standardizing standard deviation.

3, the bioassay method based on iris texture analysis according to claim 1 is characterized in that: described automatic Iris Location may further comprise the steps:

(1), definition is the center with the evaluation point, has the local assessment area Δ S that radius is analyzed yardstick;

(2), definition is contained in the mathematic expectaion and the variance of the pixel in the Δ S;

(3), by assessment area interested is produced the variance set of all points to be assessed and radius to be assessed set with interval sampling mode iterative step (1)-(2);

(4), variance sets definition to be assessed 2 rank rate of change (acceleration) maximum values to being produced in the step (3);

(5), redefine iterative parameter: with the pairing position of rate of change maximum value, 2 rank that searches in the step (4) is the regional area at center, become new assessment area interested, improve the interval sampling precision of point to be assessed and radius to be assessed, improve the analysis precision that radius is analyzed yardstick;

(6), iterative step (1)-(5), be single pixel until the interval sampling precision; Variance in the above-mentioned steps can of equal value be defined as standard deviation.

4, the bioassay method based on iris texture analysis according to claim 1, it is characterized in that: described analysis and assessment iris texture picture quality adopts and the real-time identical method of auto-focus quality evaluation, and its analyzed area is limited at effective iris region that Iris Location obtains.

5, the bioassay method based on iris texture analysis according to claim 1 is characterized in that: described living body iris is following any or its combination in any from motion tracking and detection method:

(1), living body iris detects based on dilator among the ophthalmology and expands the biological characteristic of the pupil convergent-divergent variation of about flesh generation;

(2), living body iris detects the Photobiology characteristic that changes by the mutual contrast of the brightness that produces based on living body iris biological tissue under the image-forming condition of different arrowbands emission wavelength and various combination thereof;

(3), living body iris detects the Photobiology characteristic that changes by the mutual contrast of the iris region texture that produces based on living body iris biological tissue under the image-forming condition of different arrowbands emission wavelength and various combination thereof; Said method is being implemented under the controllable emission wavelength lighting condition at random.

6, the bioassay method based on iris texture analysis according to claim 1 is characterized in that: described iris analysis geographic norms polar coordinate transform may further comprise the steps:

(1), be the 2D regional polar coordinate that standardize according to iris outer edge radius that Iris Location obtained and the coordinate transform of central coordinate of circle definition 2D card Deere;

(2), coordinate system transformation by discretization after, definition angle element sample quantity and increment;

(3), the angle element of definition is finished the 2D regional polar coordinate transform that standardizes according to the discrete sampling mode that is adaptive to different radii in the step (2);

(4), direct reflection and high quadrant eyelid, the eyelashes interference sections zone of the low quadrant of coordinate system eliminating are used to the iris texture characteristic analysis.

7, the bioassay method based on iris texture analysis according to claim 1, it is characterized in that: described demodulation is expressed iris texture characteristic and is adopted 2D m rank cardinal B-spline wavelets transfer pair, described wavelet transformation is to being the even number m rank cardinal B-spline wavelets with even symmetry (symmetry), and the wavelet transformation that the odd number m rank cardinal B-spline wavelets with odd symmetry (antisymmetry) is formed is right, and described wavelet transformation is to satisfying m 〉=3 conditional requests.

8, the bioassay method based on iris texture analysis according to claim 1, it is characterized in that: described demodulation is expressed iris texture characteristic and is adopted multiple dimensioned, the 2D Morlet wavelet transformation demodulation of multidirectional, and the direct current/DC of shielding 2D Morlet small echo real part responds output, and it makes DC response be output as 0 attribute.

9, the bioassay method based on iris texture analysis according to claim 8 is characterized in that: the design parameter relation of described 2DMorlet wavelet transformation distributes according to the requirement based on biological vision cortex model, comprising:

(1) the wavelet scale factor is according to 2 ^j(j=0,1,2 ... N) exponential relationship distributes, and the mid frequency position relation that the mid frequency demodulation factor is determined according to the corresponding wavelet scale factor distributes;

(2) the mid frequency demodulation bandwidth factor and the mid frequency demodulation factor distribute according to the constant proportionate relationship, and this constant guarantees that the band of mid frequency demodulation bandwidth is limited to and has the octave of being less than or equal to.

10, the bioassay method based on iris texture analysis according to claim 1 is characterized in that: described quantization coding produces the feature coding method and may further comprise the steps:

(1), definition is based on the broad sense binary quanta basic coding operator in iris partial analysis zone;

(2), by step (1) to all iris partial analysis area sampling parameters, produce and to possess the measurable iris texture characteristic coding of unified standard IrisID.

11, the bioassay method based on iris texture analysis according to claim 1 is characterized in that: described feature coding IrisID measuring similarity adopts the full probability method for measuring similarity.

12, the bioassay method based on iris texture analysis according to claim 11, it is characterized in that: described full probability method for measuring similarity adopts the corresponding different iris texture characteristics coding of cycle cyclic shift coding generation IrisID on the standard angle direction, it carries out repeatedly the postrotational method for measuring similarity of related angle with the registered feature coding that needs to identify or verify, final measuring similarity result is the postrotational measuring similarity maximum of above-mentioned repeatedly related angle.

13, the bioassay method based on iris texture analysis according to claim 1, it is characterized in that: the similarity decision value that described measuring similarity judgement is determined based on data statistic analysis, the individual judgement standard that it is used to determine, corresponding measuring similarity result provides acknowledge information greater than the decision value system, otherwise provides NACK messages.

14, the bioassay method based on iris texture analysis according to claim 1, it is characterized in that: the described direct reflection of removing automatically in real time is that analysis according to the imaging image obtains direct reflection degree analyzing result, adjust the position of lighting source and the incident angle between iris, reach change direct reflection degree, until obtaining the interferential iris image of high-quality areflexia.

15, a kind of based on the used iris optical imaging device of the bioassay system of iris texture analysis, it is characterized in that: it includes housing, imaging len (102) is arranged in the housing, video camera (100), and be sequentially arranged on the optical axis, it is characterized in that: the preceding vertical plane incident cold mirror (101) that is provided with of optical imaging lens (102), distance measurement sensor (104) and lighting source are set in the housing, described lighting source is adjusted the positioner (105) of lighting source position and the diffusion filter (106) of sealing lighting source and is formed by visible light-near infrared band lighting source (107).

16, iris optical imaging device according to claim 15, it is characterized in that: described visible light-near infrared band lighting source further comprises by the surface LED chip array of emission wavelength in the 400-900nm scope and the energy driver that can change the radiation of light source brightness/intensity and forming, and the synthetic and control that can be combined by the independent narrow emission wavelength that decomposes of the emission wavelength of described visible light-near infrared band lighting source.

17, iris image optical imaging device according to claim 16 is characterized in that: the selection of the emission wavelength of described visible light-near infrared band lighting source:

(1) be purpose for carrying out individual bioassay with the graphical analysis of imaging iris texture, the lighting source emission wavelength is in the 720-800nm bandwidth range;

(2) realize that for changing with the pupil convergent-divergent it is purpose that living body iris detects, the lighting source emission wavelength is in the 400-680nm bandwidth range;

(3) be purpose for realizing that with the Photobiology characteristic variations under the different image-forming conditions living body iris detects, the lighting source emission wavelength is for having that typical half-peak value ripple is wide to be different arrowbands emission wavelength and the various combination thereof of 30nm in the 400-900nm bandwidth range.

18, iris image optical imaging device according to claim 15, it is characterized in that: described vertical plane incident cold mirror is recessed surperficial cold mirror or plane cold mirror, is provided with list or eyes eyeball vision coverage area guiding sign on the cold mirror surface that possesses the function removed.

19, iris image optical imaging device according to claim 15 is characterized in that: described vertical plane incident cold mirror reflection 700nm and following wave band are used to visual feedback, and the above wave band of transmission 700nm is used to the physics imaging.