DE3943680C2 - Visually readable labeling - Google Patents

Description

The invention relates to an optically readable marking according to the preamble of claims 1 and 2, d. H. a label, which is attached to a substrate to within a certain two-dimensional data field to record and store information, where the data field consists of a number of polygons, each one of at least two different optical ones Have properties, and a method for producing them Labeling, a device for detecting and decoding the Labeling and use of the label.

Goods, individual parts, letters, packages, containers and others Items to be shipped or transported are repeated regarding the origin, the flight number, the Destination, name, price, part number and identify a lot of other information. A another application involves reading coded informa tions that are on labels or markings to automatically generate sales and inventory grasp or fed to electronic cash registers to be able to. Other applications of such coded Labels include automatic sorting and Forwarding mail, parcels, luggage and the like, and attaching labels with manufacturing instructions on raw materials or individual parts for a manufac process. The labels for such items are  conventionally provided with a bar code, for example the UPC bar code. It is quite a number of different known to those bar code systems.

However, the conventional bar code systems assign one low data density to meet storage requirements of more and more information on ever smaller labels or markings. A decrease in Size and spacing of the stripes in different Bar code systems do not solve the problem; Systems and optical scanners with sufficient resolution for the festival the provision of strips 75 µm or less apart and the required small tolerances during the printing process are just not economical. To a larger data set very large labels must therefore be used with the result that these labels are no longer on smaller items fit. Another essential fact tor are the cost of the label material, for example wise paper. The paper for a smaller label costs less than that for a large label; with large quantities these costs are a significant factor.

Known alternatives to the bar code are: Round formats with radially arranged, wedge-shaped code elements (US-PS 3 553 438) or concentric black and white bit coding te rings (U.S. Patent 3,971,917 and U.S. Patent 3,916,160); Grid off Rows and columns of coded squares or rectangles (U.S. Patent 4,286,146, British Patent 1,216,539); microscopic points in cells are arranged, which form a regular grid (US-PS 4,634,850); and densely packed multicolored data fields Dots or elements (U.S. Patent 4,488,679). Some of these games do not have sufficient data density; about the knockout dated round pattern or the grid of rectangular or square areas. The grids with microscopic points  or with multi-colored elements require a special one Orientation and promotion to be read which limits their usability accordingly.

Because of the size and speed of modern conveyor systems with, for example, conveyor belts from 90 to 120 cm wide and belt speeds of 2.5 m / sec or more that transport objects of different heights, and the need for small, cheap, compact markings or labels with one Area of approximately 6.25 cm ² (one square inch), great efforts are made in the field of capturing and reading out the coded labels on these fast moving objects. The captured or identified image of the label must be properly decoded before the item on the conveyor belt reaches the next working position, often in a split second. There is therefore a need for a simple, quick and inexpensive device for signaling the presence of a coded label in the field of view of an optical scanner which is mounted so that it can scan the entire conveyor belt. At the same time, the data should be arranged with high density.

Data fields with acquisition targets are already known for Example concentric geometric figures with rings, Squares, triangles, hexagons and variations of how in U.S. Patent Nos. 3,513,320 and 3,603,728. In the U.S. Patents 3,693,154 and 3,801,775 use Sym bolen described as identification and position indicators contain concentric circles, the sym bole attached to the objects to be optically scanned are. However, these known systems include Ver Two separate symbols are used to determine the identities tification of the data field and its position, whereby the  logical scarf required to identify the symbols tion becomes complicated and what also affects the data capacity of the associated data field reduced. At this Using two symbols will damage a sym bols to difficulties in locating the position of the data field and the associated ability to Obtaining information from the data field. Both latter systems are at opposite ends of data tracks with coded linear markings used separate position and orientation indicators, with the consequence that the ability to record data is limited.

The above systems are using an optical sensor sampled, which produces an image output signal that the Changes in the intensity of light that corresponds to Data field and the position and orientation symbols is reflected. The output signal of such systems has a certain bitmu after its digitization ster, which is compared with predetermined bit sequences. This Systems have the disadvantage, however, that two separate sym bole are required to capture the image first and then determine its orientation. Also at Process of comparing the digitized output signal les of the optical sensor with predetermined bit sequences that represent both the position and orientation symbols make, read errors easily occur, since the known characterization detection systems of the acquisition target signal level is not flexible.

In U.S. Patent 3,553,438 there is a circular data arrangement described with a centrally arranged acquisition target, that has a series of concentric circles. The he The objective of the collection is a device for acquiring the circle  shaped label through the optical sensor and the Be mood of its geometric center and thus the geometric center of the circular data arrangement This determination is made by a logic circuit, the recognizes the pulse pattern that is the circular shape of the detection represents goal. As with the bar codes, however Data field only a small data capacity, and the system requires a second circular scan. The use of a linear and circular scanning in such a sy system with limited data capacity would take a lot of effort to achieve a slight increase in the data capacity compared to conventional bar codes things; matters.

To increase the data capacity, codes with dense grabbed multicolored dots as they developed in the U.S. Patent 4,488,679. Systems of this type are required but the use of hand-held opti scanners, which makes it completely impossible to fast moving data fields on an object on a record and decode fast conveyor belt. High dense code systems with microscopic data points, as in U.S. Patent 4,634,850 require special trans port facilities to ensure that the data field exactly in a given direction, and are with any orientation, such as when transporting Ge objects on conveyor belts are typical, not applicable. The coded label must be read track by track, for which a linear scanner is provided, which with the Trans port device is connected. The position of the label with regard to the sensor, it must be observed very precisely, to read the label correctly.  

In connection with bar codes, multicolor Systems used to address the optical problems of scanning to overcome closely adjacent strips. A bar code with more than two optical properties for encoding Da alternate in a data field, for example by the en use of black, gray and white stripes in U.S. Patent 4,443,694. However, with such systems, the data density is not increased to such an extent that it meets the requirements.

The object of the invention is therefore a compact, optical create legible labels with high information density fen that is easy to read. In addition, a method for producing this label, a Device for detecting and decoding the marking and a use for storing and retrieving data can be specified.

The solution to this problem is given in claims 1, 2, 10, 11 and 12.

Due to the easy readability, the markings can be read by one optical sensor read even then if they are on an object, which is moved at high speed, the Orientation and the height of the object did not matter len should.

The optically readable marking can be decoded device can be combined, the marking reliably decoded, even if the label is tilted, curved, distorted, partially removed or partially is torn.  

The grid can be marked in the plane of two or more Have axes with the same or different angular distances.

The polygons on the label can be complete, partial wise or not related at all. The latter both arrangements result in a number of spaces between neighboring polygons. These spaces can  the same or different optical properties as the buttocks have lygone. Two-dimensional arrangements can be put together hanging polygons with five or more sides for the Labeling can be used. Can also be two-dimensional Orders regular or irregular and partial or non-contiguous polygons with three or more Pages are used when on specific axes of such a field.

In addition, the marking according to the invention can be an Er based on a number of concentric rings point, which facilitates the localization of the marking, especially with dynamic reading systems.

In a preferred embodiment, the data field includes a generally square array of 6.25 cm ² area with contiguous hexagons forming rows and columns, and with a central detection target with a geometric center that is the geometric center of the data field certainly. The detection target can have any geometric shape with optical properties that give an easily recognizable image signal when it is scanned with an optical sensor along a linear scan line that passes through the center of the detection target. The detection target preferably consists of a number of concentric rings alternating reflectivity, which give a periodic image signal when they are scanned in a straight line. By using analog filters in the detection and decoding of the data field, the signal generated by the optical sensor is compared directly with a predetermined frequency, which enables a quick and accurate checking of the frequencies and subsequent determination of the location of the data field on the substrate. The electrical analog signal output by the optical sensor, which contains the coded information in the identification, is then digitized and decoded. The use of a filtering step with an analog bandpass allows the identification to be detected without the need to decode the encoded information of the identification. A reference point on the data field can be determined by acquiring the center of the acquisition target. If the center of the detection target is in the center of the marking, the center of the detection target and the data field can be determined at the same time.

The optically readable data array of the marking according to the invention can contain up to several hundred error-protected alphanumeric characters on an area of 6.25 cm ² if hexagons with three reflection properties, for example the colors black, white and gray, are used. For a sensor with a given optical resolution, the system according to the invention allows the information to be packed much more densely than would be possible with a bar code system. If, for example, a high-resolution optical sensor is used with the system according to the invention, hundreds of alphanumeric characters can be accommodated on an area of 6.25 cm ² . On the other hand, a hundred characters can easily be detected on the same area with a relatively poorly resolving sensor.

According to the invention, optically readable markings can be made by Use two or more contrasting optical eggs properties with different data densities the. Greater data density and the provision of an acquisition solution target, however, requires a scanner with com more complicated structure and more complex decoding algorithms for reading the encoded information.  

According to the invention, the data is encoded by encoding a number of bits from a binary bit sequence into one Group of neighboring polygons, each polygon has one of at least two optical properties, although the coding is alternatively polygon by polygon could be executed. The digital bit sequence can be on the Basis of manually entered data by a computer generated or otherwise in the binary bit sequence can be changed, and one can also be recorded beforehand nete digital bit sequence can be used. The ones to be encoded Data are geometri in a certain sequence areas of the data field assigned to the number the transitions between polygons with different opti properties.

The information to be encoded is preferably given in In high and low priority formations that separated in different geometric areas of the data field. The informa high priority areas in the information areas lower priority to be duplicated to make the probability Possibility of losing the high priority information due to scanning errors due to dirt, cracks, etc. and other damage in the data field. The In High priority formations are in a central area of the data field, close to the capture that is preferably provided target, enrolled to protect the information from to protect damage that is more in the peripheral areas of the Data field occur. In the data field you can use the large information capacity of the invention Labeling options for error correction pre will see.  

To print the label with polygons of different optical properties becomes a grid of sufficient density used. Alternative printing methods can also be used find. The grid is bit-coded, so when the Marking is printed, the optical properties of each hexagon are given. It can be standard printer for printing the polygons with the required op table properties are used.

The polygons in the data field, preferably hexagons, Data contained are preferably as follows read: The coded labels go through a given illuminated area and are thereby marked by a electronically operated optical sensor or one in the Hand held scanner scanned. The optical scanner produces an output signal that is an electrical analog signal is that of the intensity of each reflective eigen area of the labeling corresponds to how it is recorded by the pixels of the optical sensor becomes. The analog signal of the op table sensor first with a predetermined frequency value compared, which ent that of a specific acquisition target speaks if this is provided in the data field. If a match has been found, the Kenn drawing captured and the center of the capture target determined, which also becomes a reference point for the data field is set. At the same time, the analog signal becomes continuous digitized in an analog / digital converter and stored in an image memory. The saved digital data representing the entire label are then available for further processing and decoding Available.  

Through logic circuits with stored programs who the digital data into an image of the interfaces the hexagons with different optical properties puts. This is preferably done by calculating the stan dard deviation of the intensities by the optical Sensor at each pixel and a predetermined group of Pixels around this were recorded. Great standard deviations correspond to transition areas on the cut represent contrasting hexagons.

The digital data is then further processed under through filter programs to determine the orientation tion, the direction and the distances of the hexagons. The we The essential process steps are:

1. Filter the nonlinear transformed version of the digitized image
2. Determine the orientation of the label, preferably by capturing the three axes of the figure and determining which axis is parallel to two sides of the label
3. Finding the center of each hexagon and determining the gray tone at the center;
4. converting the gray tones into a bit string;
5. Optionally carry out an error correction; and
6. Optional conversion of the bit sequence into a specific character set.

Embodiments for optically readable markings, their attachment to an object and their decoding are explained in more detail below with reference to the drawing. It demonstrate:  

Fig. 1 is a detection target with concentric rings;

Figure 2 shows a section of an optically readable identification drawing with adjacent hexagons.

Fig. 3 shows a complete first embodiment of an optically readable marking with adjacent hexagons with three different optical properties and with a detection target;

Fig. 4 shows a group of 3 × 3 hexagons, which serves as a basic dier unit;

Figure 5 is a graphical representation of the groups for a 33-row, 30-column data field that form a grid of 11 rows and 10 columns;

Figure 6 shows a device for adjusting the position of an optical sensor to the height of erfas to send objects.

Fig. 7 is a flowchart for the decoding process;

Fig. 8 is a flowchart for recording the acquisition target;

Figure 9 is a flow chart for the encoding and decoding program and data flow therein;

Fig. 10 is a flowchart for the steps in processing the image;

11 is a group of contiguous regular hexagons whose centers lie on a hexagonal grid PERIODIC gen.

FIG. 12 is a group of contiguous irregular hexagons whose centers lie on a hexagonal grid lar irregular shapes;

FIG. 13 is a group of partially coherent poly Gonen of substantially the shape of hexagons whose centers lie on a hexagonal lattice;

FIG. 14 is a group of contiguous polygons are of substantially the shape of hexagons whose centers are on a hexagonal lattice;

Figure 15 is a label having contiguous polygons substantially the shape of hexagons whose centers lie on a hexagonal grid, and with a detection target.

FIG. 16 is a group of non-contiguous squares whose centers lie on a hexagonal lattice;

FIG. 17 is a group of non-contiguous rectangles with gaps, whose centers lie on a hexa gonal mesh;

FIG. 18 is a group whose centers lie of noncontiguous pentagons on a hexagonal lattice;

Figure 19 is a group of contiguous rectangles which adjoin each other and whose centers lie on a hexagonal grid. and

Fig. 20 shows a group of partially connected octagons, the centers of which lie on a rectangular grid.

On the label are by means of the contrasting color information from neighboring hexagons or cells th, which are detected by an electro-optical sensor can.

Polygon cells other than hexagons can also be used if the geometric centers of adjacent polygons lie on the corners of a hexagonal or other predetermined grid. In Figs. 16 to 18 are the co decode units such polygonal cells by a closed dashed line. The cells are always arranged in a certain pattern. A wide variety of polygons can be used, and grids of various geometries, such as hexagonal, rectangular or square grids, can be used. "Adjacent" polygon cells can be connected, partly connected or not connected at all.

"Connected" are polygons whose boundaries are different touch along the entire periphery so that none Spaces result. "Partially connected" polygons are arranged so that the boundaries along the Touch the periphery in places and not in places stir, creating a number of spaces between the Polygons are created. "Non-contiguous" are polygons who have no contact at all along their boundaries the surrounding polygons.

The polygon cells and the grids on which the funds are located points of the polygons can be irregular with unevenly spaced axes, or regularly, with equal spaced axes. The grid axes can be independent of the axes of symmetry, if any, of the polygon cells.

Hexagons have some for coding information essential advantages, especially the following:

• 1. For a given optical resolution, hexagons are packed more densely than other polygons. For example are the corners of squares for a given resolution difficult to recognize, so that for "reading" squares a in itself unnecessary optical resolution is required. Circles the gap would be ideal for the optical resolution between neighboring circles, however, is wasted and complicates the manufacture and printing of the labeling, because the gaps have their own optical properties order is.  
• 2. A grid of neighboring hexagons has three axes. When labeling with square or rectangular The main axis of the hexagons can be predetermined by a shape bene relationship to one side of the label slightly lo be calibrated. This localization of the major axis of a Hexagonal grids make it easier to read the encoded Data.

A separate unit is to be found here under "Identification" stand by means of a suitable adhesive back a package or product, the outside of a container or can be attached to another object and on which optically readable information is printed.

With "optically readable data field" or "data field" for short a pattern of adjacent hexagons or cells with two or more optical properties, that is serves, in a recoverable form, a data set using the respective optical properties and the mutual to represent the spatial relationship of the cells.

The pattern of adjacent hexagons with a maxi paint number of transitions from hexagon to hexagon to optimal readout and for a maximum storage density referred to as "honeycomb pattern".

The contrasting reflective properties of each Hexagons or cells of the data area can refer to the ver various types are generated. The term "print" has the general meaning of applying Ma material with certain optical properties on one Substrate or changing optical properties, such as it for example in the so-called "thermal" printing of the Case is. The term "printing" also includes an omission  the application of a material with a certain optical Property on an area of the substrate that itself has a different optical property. To the Forming black and white hexagonal cells For example, it only required the black cells print when the substrate is white.

Under "optical properties" are the light absorption, the reflection and / or refractive properties of the cells understand. If cells are black (dense black printing) be) and gray (halftones of black) are printed as well as white (no printing on a white substrate), the identifier has drawing three different optical properties.

As shown in Fig. 1, a "number of con centric rings" 10 are to be understood as two or more concentric rings 12, one of which is formed by a circular area 15 defined by the smallest radius "r" of the rings is.

Fig. 2 shows a section of an electro-optically scannable marking according to the present embodiment. The label contains a number of adjacent hexagonal cells, which are arranged in a honeycomb pattern. Each of the individual hexagons is designated by the reference symbol 20 and has six identical sides 22 . The internal angles "a" of the hexagons are also of the same size and are each 120 degrees. The hexagons have a long vertical axis yy and a shorter horizontal axis xx.

A tag 30 ( FIG. 3) measuring approximately 2.5 cm by 2.5 cm (one inch by one inch) contains approximately 888 hexagons or cells 20 if it is considered that a detection target 35 extends around the center of the label a number of concentric rings is provided. The adjacent hexagons 20 form horizontal lines "R", which are determined by imaginary lines 31 , and vertical columns "C", which are defined by imaginary lines 33 . In this example, the label has a total of 33 horizontal rows "R" and 30 vertical columns "C". Each individual hexagon has a "diameter" of approximately 0.8 mm. Due to the geometrical packing of the hexagons, there are more rows "R" than columns "C" in a square outline of a honeycomb pattern.

As can be seen from FIG. 2, the hexagons are arranged in vertically offset and overlapping vertical columns with alternating vertically spaced hexagons with collinear yy axes. The yy axes of spaced hexagons 20 are aligned with the outer vertical sides 22 of the interposed, displaced hexagons. The yy axes of hexagons 20 are in the representation of FIG. 3 to the two vertical parallel boundaries 32 and 34 of the marking. The horizontal lines "R" are measured by the xx axes at the center of the hexagons 20 .

The hexagons 20 are ausgebil det by a printing operation, in which the hexagons 20 in two or more different optical properties such as contrasting colors are printed. These colors can be the colors white and black, so that white hexagons or cells 25 and black cells 26 are formed, as shown in FIG. 2, and gray can also be used as a further color, so that gray cells 27 are formed as shown in FIG. 3. The color tones white, black and gray were chosen in the present embodiment because they provide an optimal contrast, which facilitates identification by an electro-optical sensor. The gray tone is chosen so that its optical properties lie approximately in the middle between the optical properties of white and black color.

The label of FIG. 3 may represent a separate label that has an area of 6.25 cm ² (one square inch). Alternatively, if there is an acceptable color background (preferably white), the marking can also be printed directly on the surface of an object. However, because of the importance of a defined background for the optical properties of the contrasting colors, it is often more advantageous to use a separate label because its background color is easier to control.

The alignment of the hexagons printed on the label with respect to the sides of the label is important for the later determination of the main axis of the label. The label is printed so that the yy-axes of the hexagons in the illustration in FIG. 3 run parallel to the vertical sides 32 and 34 of the label.

For "reading" the hexagonal arrangement it is for decoration important information, a sharp color contrast to have between neighboring hexagons. The scan direction and the software required for decoding the less optical properties are used, the easier it will be Coding of the hexagons can be used. Less optical However, properties also reduce the data density of the Labelling. As a compromise between the amount of encoded Information that is stored on the label can, and the effort for scanning multicolored Kenn The use of three different drawings optical properties, especially the colors black,  Gray and white, found to be advantageous.

The gray hexagonal cells are printed on the Cells with black ink in every fifth pixel of the grid of a dot matrix printer using a halftone algorithm as all is commonly known. The printer thus prints a specific one th percentage of the pixels to determine a given gray hexagon while printing a black hexagon every pixel of this hexagon is printed out. The halftone algorithm used has the program drawing "LABEL".

The black hexagonal cells can be printed using an ordinary black ink. The Scan analysis software for decoding the license plate makes a clear distinction between the reflectivity of Black, gray and white, so that no particularly accurate feast laying of the color is necessary. However, if other colors are used, or if different shades of Gray used to be four or five color data fields to produce, the contrast of the printing inks must be more precise be trolled to measurable differences in the optical Ensure properties of different colors. The However, using a black color is the easiest and easiest way of creating a honeycomb arrangement of hexagonal cells with three different opti properties.

Due to the square shape of the marking and the nature of the hexagonal cells, there are incomplete hexagons 56 on the edges of the weave structure; 3 as seen from the Fig., these incomplete hexagons are not used to, are hereby requested any useful information.

The present embodiment of the label also includes an acquisition target. The detection target 35 shown in FIG. 3 consists of a number of concentric rings of contrasting colors (shown as black and white). The black rings are designated with the reference characters 42 , 46 and 48 and the white rings with the reference characters 44 , 50 and 52 . The detection target is preferably located in the geometric center of the marking in order to reduce the likelihood of damage or destruction if the plate is scratched, soiled or damaged at the edge. Also, the size of the image memory required to store the data before the detection target is identified is smallest when the detection target is in the middle of the label.

The number of concentric rings for the acquisition target is any, but it has been found that six con centric rings are sufficient.

To fit a calculated pattern to what is expected becomes what the concentric rings represent if that Pattern is read using a pattern correlation method used. In the case of a match, the goal is the acquisition localized, as explained in more detail below becomes. The generated in the present embodiment and The special filter used has the file name "FIND.C".  

The detection target has an overall diameter that is smaller than the data field and comprises an area of up to 25%, preferably about 7%, of the area of the data field. The detection target should be as small as possible, since the area occupied by it on the label cannot contain any information. In the present embodiment, the outer ring 52 has an outer diameter of approximately 7.45 mm. In the labeling of FIG. 3, the detection target 35 thus occupies about 7% of the area of the marking 30 of 6.25 cm ^2. Thus, the amount of information that can be contained in the data field surrounding the acquisition target is not excessively limited. As with the incomplete hexagons on the outer edge of the marking 30 , the incomplete hexagons 55 around the limitation of the detection target 35 are not used for the purpose of encoding information. The width of each ring is preferably approximately equal to the width of the hexagons (length of the xx-axis in FIG. 1) in order to facilitate the resolution. Six rings are sufficient so that the detection target can be localized with the smallest possible area and a minimum of possible incorrect readings due to "disturbing" signs on the number plate or other "disturbing" signs outside the number plate, for example on a conveyor belt.

The acquisition target can also have a different form than that of have concentric rings. For example, squares, Spirals or hexagons are used to make transitions between contrasting concentric figures as long as cuts are being made through the target clear, predetermined and identifiable color transitions ben that are suitable by an electro-optical sensor detected and measured using a suitable filter become. It should be noted that a spiral does not have an open  succession of concentric rings represents that ever depending on the size and radius of the spira le a good approximation to the shape of concentric rings can be reached. A capture target from concentric Wrestling is convenient because the signal that comes with a scan through the center of which is generated has a frequency that is always the same regardless of the direction of the cut. This makes it easier to identify the center, and it allows identification of the location of the acquisition target with a one-dimensional search of the analog or digital output signals from the scanner. However, it can alternatively or subsequently also a two-dimensional one Search for the purpose of increased accuracy, when analyzing a digital signal.

Under "concentric rings" are complete rings, part rings in the form of semicircles, sectors more concentric Rings that are between 180 degrees and 360 degrees, and Spira len, which form approximately concentric rings.

Since each hexagon can be encoded in accordance with three different optical properties in the present embodiment, 1,585 "bits" (log ₂ 3) of information can be contained in each hexagon. If fewer or more optical properties than three are used, the number of bits that can be accommodated in a hexagon changes accordingly. The encoding algorithm is designed to approximate the maximum data density and to maximize the number of transitions in cell-to-cell optical properties to facilitate the two-dimensional clock recovery process described below .

FIG. 4 shows a group of 3 × 3 cells, that is to say nine hexagonal cells 60 , of the basic coding unit which is used in the present embodiment. This is one of many possible, essentially equivalent coding forms. The group of 3 × 3 cells from hexagons 60 is intended to encode 13 information bits if the group contains nine complete hexagons, or to represent less than 13 bits if the group is incomplete due to unusable hexagons. Around 1292 bits of information can be accommodated on a marking with an area of 6.25 cm ² and a data field with around 888 hexagons and a detection target of around 7% of the marking area.

When coding each group, the outer lower hexagons 62 and 64 of each group 60 , as shown in FIG. 4, are set in their respective optical properties so that they always differ from the intervening hexagon 66 . The hexagons 62 and 64 can therefore only ever correspond to one bit per hexagon. In this way, it is possible to accommodate 13 bits for information by coding 11 bits in the remaining seven hexagons. Since seven hexagons give more combinations than are used (3 ⁷ = 2187 combinations compared to 2 ¹¹ = 2048 combinations), certain combinations are not used, e.g. all cells black, all cells gray and all cells white or all in essential black, gray and white combinations. The reason for the need for contrasting colors for the hexagons 62 and 64 relative to the hexagon 66 is to ensure a transition required for the clock recovery step and an optional normalization process and to facilitate the determination of a horizontal alignment of the data field .

If the codable group has seven or eight hexagons around the seven usable hexagons with 11 bits encoded and the eighth hexagon, if it exists, with one bit. For all other subgroups, each pair of hexagons three bits and each remaining individual Hex bit assigned.

It can be seen that this marking is a particularly effective marking, which is easy to read by means of a suitable scanner and the corresponding analysis software, with a very high density of information on a relatively cheap, easy-to-print label or the like. As described, in the present embodiment, a pack of hexagons with 33 rows and 30 columns is used on an area of 6.25 cm ² and with a detection target which occupies approximately 7% of the total area of the license plate. A group of nine hexagons contains 13 information bits, so that 1.44 bits are recorded per cell. This is less than the theoretical 1.585 bits per cell due to the other limitations of the coding algorithm, since not all 37 patterns are used and because some of the at least optically possible transitions from cell to cell are not used.

It is convenient to have a degree of error protection at the Provide coding of the labeling so that the actual quantity on the label, again obtainable information in favor of high data security is reduced.

The above description of a label with hexago nalen cells is directly on a label with others Transferable polygon cells. When using the optical egg  black and white and optionally gray also apply for other polygon cells, for example wise about data density. As with the markings Hexagons can mark with other polygon cells read out with simpler scanners if only two optical properties for encoding the information be used.

The processes for encoding the information and the algo rithms for markings with hexagons are directly on Labels can be transferred with other polygon cells. How the hexagon markings are not complete Polygon cells at the edge of the marking and around the detection target is not used to encode information.

A "honeycomb pattern" comprises an array of hexagons together menhängenden 310 (Fig. 11), the geometric centers 311 lying at the vertices or intersections 311 A of a hexagonal lattice. Regular hexagons, i.e. hexagons with six identical sides and six identical internal angles form a hexagonal grid, which in turn is regular and has three axes (A1, A2, A3), which are at an angle of 60 degrees to each other.

If the hexagons 320 ( FIG. 12) of the marking are not regular, but are symmetrical, i.e. if they are pulled apart along two parallel sides 321 and 322, for example, the geometric centers 325 of adjacent hexagons again describe an uneven hexagonal grid 327 . Such an uneven hexagonal lattice still has three axes (A1, A2, A3), which, however, no longer have a uniform angular spacing between them.

Although the hexagonal grid of Fig. 12 is not uniform, it is still a two-dimensional geometric grid with axes at predetermined angular intervals. The positions and distances of the centers of the hexagons on the axes are also specified. The geometry of the hexagonal grid is used for decoding during the decoding process. In particular, the filtering, which is carried out on the transformed digital data, which correspond to the image captured by the optical sensor, is based on the predetermined geometry of the identification. The reconstruction of the image also provides missing grid points. Missing grid points result from the fact that there are no optical transitions between polygons with the same optical properties.

In the case of uneven hexagonal grids of the type shown in FIG. 12, it is advantageous to base the main axis determination step 3 e of FIG. 7, which is carried out after a Fourier transformation for identifying the main axis of the identification. The centers of the polygons lying on this axis have a different spacing on the main axis of the marking than on the other two axes.

Refinements of the marking are possible which approximate the preferred embodiment with hexagonal cells using certain polygon cells. Fig. 13 shows the structure of a marking with polygonal cells 330 which substantially form hexagons, however, consist of 20-page polygons. The polygons can of course have more or less than 20 sides. The polygons only partially adjoin one another, in contrast to the imaginary hexagonal cells 331 in which they are drawn and which are completely connected.

The spaces 332 of FIG. 13 may or may not have a different property than the encoded polygons. Spaces do not contain any encoded information, so their presence results in lower data density for a given optical resolution and performance. If the spaces also have a different optical property than the adjacent polygons, more optical transitions are detected by the optical sensor, and the clock signal energy in the transformation area of the decoding process is therefore higher.

Since the polygons of FIG. 13 are arranged on a hexagonal grid 335 with three equally spaced axes A1, A2, A3, the geometric centers 333 of the polygon cells 330 lie on the corners of a hexagonal grid 335 . The distances, positions and spatial orientations of the center points of the polygons are predetermined and can be determined in the transformation area of the decoding process.

In the arrangement of FIG. 13, polygons are used which have almost the shape of hexagons. Since they approach hexagons fairly well, an optical sensor with a medium resolution will "read" them as hexagons.

FIG. 14 shows an arrangement similar to FIG. 13, but the polygons 340 of FIG. 14 are arranged so that they are completely connected. These polygons 340 can be approximated by an imaginary hexagon 341 , but there are no gaps between the polygons. Such a coherent arrangement is inexpensive to simplify the decoration process, but is not absolutely necessary.

The geometric centers 342 of the polygons 340 are again on a hexagonal grid 345 , have essentially the shape of hexagons and appear as such with a medium optical resolution.

Fig. 15 shows an enlarged section of a label as it arises when it is printed with a matrix printer with 200 pixels per inch (2.54 cm). The polygons 360 of FIG. 15 illustrate the shape of the geometric figure which, due to a low pixel density, is actually printed out instead of a hexagon. Printers with a higher pixel density result in a better approximation to the hexagon. The polygons 340 and 360 of FIGS. 14 and 15 therefore correspond to the shapes which are practical when the marking is produced with some printers. In practice, however, polygons of such shapes are also equivalent to cells made of connected hexagons.

As in the case of FIG. 3, the identification of FIG. 15 also contains a detection target 370 from a number of concentric rings 371 to 376 . Like the hexagons of FIG. 3, the polygons 360 of FIG. 15 are arranged in columns "C" and rows "R", the boundaries of which are formed by the imaginary lines 361 and 362 or 362 and 364 . The center points of the polygons 360 of FIG. 15 also lie on a hexagonal grid which is spanned by axes A1, A2 and A3 which are equally spaced.

If another marking geometry with a quadra table or rectangular grid is used, the two-dimensional clock recovery process, which in fol is still described, adjusted accordingly. The other geometry of the given grid makes it  required that the filters used in the filtering be changed. The filters act on the transformed ones digital data that matches the optical properties of the Correspond to polygons in the image area from the sensor were read. These relatively minor changes to the Filtering can be carried out easily by any specialist become. If the given two-dimensional grid is not has evenly spaced axes or irregular in structure is moderate, it can be favorable to the main axis of the Kenn drawing before executing the Fourier transform tify, because then the geometric centers along the Axes are not evenly spaced.

Polygons which are not arranged in a continuous manner can also be used to generate an optically readable marking. Fig. 16 shows a hexagonal arrangement of Qua draten 420, not related and their geometric centers 422 lying respectively on the vertices of a hexagonal lattice obtained by the three equally-spaced axes A1, A2 and A3 is spanned. This configuration clearly represents a grid of imaginary hexagons 421 that can be placed over the polygons 420 , with spaces 425 being formed.

An arrangement similar to FIG. 16 can be constructed using rectangles. FIG. 17 shows a number of rectangles 430 , the geometric centers of neighboring rectangles lying on the corners of a hexagonal grid that is spanned by the intersecting axes A1, A2 and A3. Recognizing the hexagonal arrangement is again made easier by the imaginary hexagons 431 , which can be placed over the non-contiguous rectangles 430 , with spaces 435 being created between the rectangles 430 . Fig. 18 also shows a non-coherent designation of pentagons 440 , the geometric centers 442 adjacent pentagons 440 lying on the three equally spaced axes A1, A2 and A3. The geometry of the non-contiguous pentagons is again more easily recognized if imaginary hexagons 441 are placed over the pentagons 440 , gaps 445 being formed.

Other hexagonal grids can also be constructed in which the axes A1, A2 and A3 have a uniform angular spacing, but do not correspond to the axes of symmetry of the polygons. Only the geometric centers of neighboring polygons lie on the intersecting axes of the arrangement. Such an arrangement is shown in FIG. 19, it comprises a number of connected rectangles 450 whose geometric centers 451 lie on the axes A1, A2 and A3.

Higher order polygons can also be arranged on a predetermined two-dimensional grid. Fig. 20 shows a number of partially contiguous octagons 460 defining a plurality of interstices 461st The centers 462 of adjacent octagons 460 are located on the intersecting axes A1 and A2. The spaces 461 can have a different optical property than the octagons 460 . However, this is not absolutely necessary since it is the position, orientation and intensity of the optical properties in the center of the octagons 460 , which are at a predetermined position in the hexagonal grid spanned by the axes A1 and A2, that are most important for the decoding process are.

In the preferred embodiment, the label has an area of 6.25 cm ² to accommodate an acceptable amount of data for 100 alphanumeric characters with a high level of error protection on a relatively small label. The use of four, five or more different optical properties or colors for the hexagons allows light to accommodate much more information in the same space, but with increased demands on the software and the scanning system for the recovery of the information. For practical purposes, a coding system with three optical properties, the colors black, gray and white, has proven to be very cheap.

The cell groups do not necessarily have to be 3 × 3 Cells exist, it can be any other shape for the group pen can be used. Any group formation is possible be omitted and the coding algorithm directly on a Patterns made up of individual hexagons.

The following is the coding process in the present Embodiment explained.

At the beginning of the coding process there is a certain series of Data that are coded on the label should. It is believed that the labeling for the Shipping is provided and the data is divided into two fields split up as "high priority information" and "Low priority information". More than two priority levels can also be provided his.

If the label is intended for shipping, you can the high priority information is nine characters long for example, the zip code of the recipient's address  Contains item, package or letter. Nine characters are therefore envisaged, since up to nine digits internationally ge postal codes are used. When sending counter the postcode provides the most important information It sets out the general destination of the item and makes it possible for scanning and sorting systems to Item on a conveyor belt or the like to that bring the right truck, plane, etc.

For example, the low priority information the name and shipping address, including the mail guide number, the recipient of the item and information included about invoicing.

The reason for generating information is high and never The third priority is to protect information of a high level Priority with special error correction options and around the high priority information more in the middle of the Labeling where the probability damage or destruction is less, and to it to allow the high priority information in the Low priority information is repeated so that, even if the high priority information is selectively broken down are disturbed, there is a possibility the information is higher Recover priority from those lower priority. When placing the high priority information in the middle In the area of labeling it is possible to agree to some Purposes of only decoding this information that only part of the data needs to be processed, resulting in a higher processing speed. This is the case, for example, when a package is on a conveyor belt and only determines the postcode needs to be to continue the path of the package on a Control number of conveyor belts.  

Due to the lower priority are the corresponding ones Information on the label is not duplicated. Both the information high and the low priority However, various error protection codes and cor rectification options include the probability an exact recovery of both types of information to lift.

By using error protection signs as part of the encoded information can be created using a suitable program and a suitable computer the system during decoding Detect and correct errors. Like error protection codes is generally known.

The responsible for the production of a label The operator can manually fill in the required data Enter a suitable computer terminal, whereupon the computer activated a printer, the label or the label with the information contained in the hexagons higher and print lower priority. The information don't necessarily have to be in such higher and lower Priority can be divided, and is also a capture target not absolutely necessary, as is the case with the present Embodiment in the middle of the marking with an An number of concentric rings made of two contrasting rings Color is provided, the colors being two of the colors exist that are also provided for the individual hexagons are.

The operator manually inputting the data causes the suitably programmed computer to encode each character entered and use appropriate field identifiers to generate a binary bit string that represents the characters of the information and is appropriately encoded to provide high and low priority information and determine the relative position thereof. This operation is performed by the program "TEXTIN.C", denoted in FIG. 9 by the reference numeral 110.

Alternatively, the process can begin with the coding information already as a binary bit sequence are present, which are obtained, for example, from a memory will be or otherwise generated.

After the binary bit sequence has been generated or after an error-protected bit sequence has been produced as described below, the bit sequence must be implemented in accordance with a specific pattern for the encoding in the hexagonal honeycomb form. Fig. 5 is an illustration of the groups showing the individual hexagonal cells of 3 × 3 cell groups arranged as a grid or honeycomb with 33 rows and 30 columns. Each row and column is numbered with row numbers from 1 to 33 or column numbers from 1 to 30. Some of the hexagons on the right edge of the group card and in the middle of the grid are marked with an X. This indicates that these hexagons contain no encoded information. The characters X on the edge represent incomplete hexagons, which means that each of the corresponding lines contains one less hexagon. The characters X in the middle stand for the spaces occupied by the acquisition target or the incomplete hexagons around the acquisition target. All hexagons that are not marked with an X can include information. In the present embodiment, each of these spaces is occupied by a black (B), white (W) or gray (G) hexagon. Although, as mentioned, a wide variety of grouping and assignment techniques can be used, groups of nine hexagons with three rows of three hexagons each are provided for specifying certain information bits, and 13 information bits can be recorded in each such group with nine hexagons.

In a data field with 33 rows and 30 columns of adjacent hexagons, a grid of 11 rows and 10 columns of hexagon groups can be formed, each containing 3 × 3 adjacent hexagons, as shown in FIG. 5. However, it is apparent that each row of the groups of 3 × 3 cells in the 11 × 10 group grid contains a group of either seven or eight hexagons due to the geometrical packing of the hexagons, and their number varies from row to row. This arrangement thus results in six groups with eight hexagons and five groups with seven hexagons. The acquisition target accommodated in the middle also creates incomplete groups. Fig. 5 is thus a graphical representation of the usable groups of hexagons that are available in a data field of 33 rows by 30 columns for encoding information.

Groups with nine usable hexagons are encoded using the following algorithm (see Fig. 4):
Extracting 11 information bits and assigning them to the set of seven hexagons labeled a, b, c, d, e, f and h; and use of the hexagons g and i to represent one bit each such that each of these hexagons differs from the hexagon h. 13 information bits can thus be represented by a complete group of 3 × 3 cells with nine adjacent hexagons.

For groups with seven or eight usable hexagons:
Extraction of 11 information bits and assignment to the first seven usable hexagons; the eighth hexagon, if present, is used to represent a bit.

For all other incomplete cells:
Assign three bits of information to as many pairs of hexagons as possible. Use of remaining individual hexagons to represent a bit.

Since the coding of seven hexagons gives more possible combinations than corresponds to 11 bits (3⁷ = 2187 versus over 2¹¹ = 2048), some combinations of hexagons are not provided. These non-intended combinations are chosen so that they are the ones that give the least number of transitions. To do this, look-up tables are set up to form the groups as shown in FIG. 5. The creation and use of such lookup tables is familiar to every programmer. The programs for generating the look-up tables "BINHEX.LUT" 132 and "HEXBIN.LUT" 134 shown in FIG. 9 are contained in a program "MK HEX LUT" 130 .

The use of such a bit allocation scheme enables light the placement of 1292 bits of information in one Data field of 33 rows and 30 columns side by side gender hexagons.  

The order in which the information is high priority and the low priority information in the group card will be arranged depending on

• (a) the level of high priority information;
• b) the level of low priority information; and
• c) the optimal placement of the information high Priority in a protected area certainly.

Using the group map of Fig. 5 as a model, a stored mapping program "MKMAPS.C" 140 predetermines, on the digital data stored in a memory, how the information - both high and low priority - is supplied to the group map are distributed.

In order to keep the probability of errors occurring as low as possible and to be able to correct errors, error protection and error correction options are provided. For example, when coding 1292 information bits in a 6.25 cm ² area with 33 rows and 30 columns of hexagons and a detection target that occupies about 7% of the area, it is favorable to use 36 information bits of high priority, to encode a nine-digit zip code and an additional alphanumeric character that can represent a shipping code. In this example, it is advantageous to use 120 check bits for the high priority information, as dictated by the extent of the desired error correction option. 560 bits are provided for the low priority information; it includes 40 bits for repeating the high priority information in the low priority information. In addition, 576 check bits are then provided for the low priority information in order to obtain the necessary security and the possibility of recovering the low priority information. This example illustrates the much wasteful use of check bits for the high priority information than for the low priority. The example mentioned is only for explanation, and the number of information bits for the different priorities of the check bits etc. can be chosen larger or smaller than specified depending on the respective requirements.

A "systematic code" has a certain information follow and add a separate error to the message sequence test sequence added. A "non-systematic" code just has if a certain message sequence and includes to together with the message sequence the error check sequence, so that the message is no longer separated from it, but again is recoverable. It can be both a systematic and a non-systematic code used for error protection become. The following description includes usage a systematic code.

The step of "inserting error detection symbols" can be systematic and / or non-systematic coding system systems include.

There are a number of systematic, linear and cyclical ones Error protection codes known, for example BCH codes, Reed-So lomon codes and Hamming codes. In the present version form are separate Reed-Solomon codes inserted in order to Completeness of information high and low priority  ensure. Reed-Solomon codes are very much us and are mostly used when multi-bit time should be checked for errors. Reed Solomon codes are generally known. However, it can also be any other type error correction codes. Reed-Solomon-Ko des and other coding systems are described, for example, in "Theory and Practice of Error Control Codes "by Richard E. Blahut, Addison Wesley 1983, described on pages 174 and 175.

Some details of the Reed-Solomon-Ko of the example. Certain properties of the Reed-Solomon codes correspond to the following parameters:

m = number of bits in each symbol;
n = number of symbols in a block = 2 ^m -1;
k = number of information or message symbols (number of information bits = km);
t = possibility of correction in the number of symbols = (n - k) / 2.

A nine-digit zip code and a single alphanume Require sign for further identification purposes 36 bits in the above example without error protection. For the A Reed-Solomon code became high priority information selected with the following parameters:

m = 6 (6 symbol bits);
n = 2 ⁶ -1 = 63;
t = 10;
thus k = n-2t = 43.

Since only six 6-bit symbols are required to represent a 36-bit message, 37 symbols ( 43-6 ) remain as padding symbols, which are contained in the encoder and decoder and need not be indicated on the label. The total number of bits required for the high priority information on the label is therefore equal to (63-37) × 6 or 156 bits.

With this error correction scheme it is possible to a maximum of to correct 60 (10 × 6) bit errors, which is 38.5% of bits used. Due to the large number of included padding symbols it makes the big bug-proof position capacity of this Reed-Solomon coding extremely un likely that the high priority information is wrong be read out.

The low priority information is Solomon code encoded with other parameters, namely with

m = 8 (8 symbol bits);
n = 2 ⁸ -1 = 255;
t = 36;
k = n - 2t = 183.

Because in the example described on the marking 1292 Bit coded can be specified, a total of 1136 bits (1292-156 bits for high priority information and check bits) for coding and for the check bits of the information low priority available. The remaining 904 bits (255 × 8-1136) should therefore be provided as padding bits. The gives 560 bits (183 × 8-904) for the information low priority and 576 check bits.

To regain high priority information further secure, these are lower in the information Priority included again. The one on the information low-priority Reed-Solomon error protection  code enables the coding of an additional 86 alphanumes 6-bit characters with maximum error correction ability of about 25.4%.

With the Reed-Solomon error protection coding above, the Total number of 1292 bits of information shown on the Labeling are available, distributed as follows:

36 bits for high priority information;
120 check bits for this information;
560 bits for low priority information (including 40 bits for repetition of high priority information); and
576 check bits for the low priority information.

The bit sequence of the data, including the corresponding check bits to ensure the information, are assigned to the individual hexagons in the group map of FIG. 5 as shown. A large number of distribution patterns can be used if the important criteria to be determined are taken into account:

1. Secure arrangement of the high-priority information in the vicinity of the acquisition target (if such is provided in the data field) or the center of the identifier; and
2. Generation of a pattern that can be easily recognized when reading.

The error used in the present embodiment coding program is "ERRCODE.C".  

The coding for the Reed-Solomon codes requires a multiplication of the information code vector with a generator matrix. This matrix multiplication is carried out by means of a Ga lois field arithmetic. The addition of two elements of the field is carried out by performing an exclusive-OR operation between the two elements. A multiplication takes place via a "log" operation in the Galois field. The log and antilog values are obtained using look-up tables derived from the main polynomials, especially 1 + x ⁶ for the high priority information and 1 + x ² + x ³ + x ⁴ + x ⁸ for the low information priority. An auxiliary program "GF.C" 126 ( FIG. 9) generates the look-up tables required for the Galois field arithmetic. The lookup tables are calculated and stored in the "GF.LUT" file 127 for use in coding and decoding. The generator polynomial g (x) for the Reed-Solomon code is determined using the following equation:

g (x) = (x + a) (x + a ² ). . . . (x + a ^2t );

where a is the basic element of the Galois field.

The generator matrix for the Reed-Solomon code is formed by performing a longitudinal division for each of the rows of the generator matrix. The kth line of the generator matrix is formed from the remainder that is obtained by executing the longitudinal division of x ^nki by g (x).

The calculation of the generator polynomials g (x) and the generator matrices for the information of high and low priority is carried out according to the utility program "MKRSLUT.C" 125 .

The lookup tables for the generator matrices are stored in the file "RS.LUT" 128 .

The markings with the hexagons can be produced by means of standard printers which are available everywhere and are cheap. A printer with a matrix of 300 × 300 dots per square inch (6.25 cm ² ) is sufficient for printing three-color (black, gray, white) markings with 888 hexagons plus a central acquisition target. Such a printer is, for example, the Hewlett Packard Laser Jet Series 11 model with a 0.5 megabyte memory and a graphic resolution of 300 dots per inch (48 dots per centimeter). A grid of 300 × 300 pixels with a density of 90,000 pixels per square inch gives approximately 90 pixels per hexagon. A value of 0 or 1 is assigned to each pixel, depending on whether the pixel represents black or white. The printer can be used to print a two-color data field from black and white hexagons or a tri-color data field from black, white and gray hexagons if a halftone algorithm is used to generate the gray hexagons.

As shown in FIG. 9, an area lookup table "REGIONS. LUT" 141 with 34 rows and 30 columns is generated by means of a program "MKMAPS.C" 140 , which is analogous to the illustration of FIG. 5, but beyond that because it is intended to specify the selection of black or white for the target rings. The individual hexagons are encoded in black, white or gray or cannot be used. Through a stored subroutine of the program "MKMAPS.C", which defines the affiliation of each of the 300 × 300 pixels in the grid to specific areas in "REGIONS.LUT" 141 , ie approximately 90 pixels per hexagon, a separate lookup table "HEX" is created MAP.LUT " 142 generated. The pixels belonging to the detection rings are encoded either in black or white. The detection target rings are printed by first forming a hexagonal pattern for each area line and then creating the rings. The areas partially or completely covered by the acquisition target rings are specified in "REGIONS.LUT" 141 as unusable.

The error-protected, coded bit sequence is transmitted into the field of 11 × 10 groups of hexagons according to a predetermined sequence. The sequence is determined by a look-up table "ORDER.LUT" 151 , which is generated by a stored auxiliary program "ORDER.C" 150 . The program "PRLABEL.C" 160 assigns values 0, 1 or 2 to areas which are available for printing on the marking, while areas with a value of 3 are left unchanged. Gray tones for the hexagons in the groups of 3 × 3 cells are assigned in connection with a program with the title "CELL CODE.C" 170 .

This utility program provides that the information of high priority is preferably stored in an area around the detection target in which the likelihood of damage to the identification is less. The program "LABEL.C" 180 is used to generate a bit sequence which is suitable for input into the laser printer.

Printing labels in black, gray and white is an easy process since only black printing Color is required if an ordinary halftone algo rhythm is used. If other color combinations are featured see what is easily possible are appropriate more complex printers required.

If a black or white value has been assigned to each of the printer's pixels, the labels can be printed accordingly to produce an encoded area as shown in Figure 3, in which some hexagons are white, others are gray, and again others are black and in which a detection target is formed from concentric black and white rings in the center.

The following is the process of decoding or off evaluating the license plate explained. It is obvious, that the evaluation preferably with very high speed on the order of a fraction of a second done to the effectiveness of handling the item (or other manipulation).

There are basically two ways to capture the image when reading the label. The label can be read at a relatively slow rate using a handheld static fixed focus scanner. Alternatively, an electro-optical sensor with a servo-controlled focusing mechanism can be used for dynamic, fast scanning of fast-moving objects of different sizes and heights. The method described below and the corresponding device for decoding are explained in conjunction with the use of a fixed focus scanner. However, the process can also be transferred to a dynamic scanning system with certain modifications in the optical system. For handling objects at high speeds, a high-speed scanning mechanism is preferably used, which can read markings that pass a fixed scanner position at a linear speed of about 2.5 m / sec or more. Fig. 7 shows the sequence of the steps of the decoding process, and the image processing includes the steps described below.

When an object, package or letter is transported on a fast conveyor belt, the area to be illuminated is relatively large and the dimensions of the objects on the conveyor belt are different. For example, 1 m wide conveyor belts and packages with a width of a few centimeters up to almost a meter and corresponding heights are not uncommon for package distribution systems. The marking with an area of 6.25 cm ² can be anywhere on the entire width of the conveyor belt. The packages can also be inclined with respect to the direction of movement. The parcels, letters and the like have different heights, so that the markings to be scanned can be located only a few centimeters above the conveyor belt or almost one meter above the conveyor belt, corresponding to the maximum height provided by the respective system.

To adequately illuminate the markings, see below Consideration of the wide range of possible packages widths and heights and at possible display angles Using an intense light source is beneficial, which optical properties chosen for labeling were reflected, defined. The light can be in the infrared, ultraviolet or visible wavelength range and at  for example in the visible range of the spectrum at any lying place.

The light source for the lighting must therefore be strong enough to give sufficient reflected light at the light sensor (z. B. a CCD element) so that the sensor reliably between the black, gray and white or other optical properties of the hexagons used can divorce. In the case of a dynamic scanning system, an LED array can be used to obtain an illumination of approximately 10 mW / cm ² at the height of the marking. The LEDs can be arranged flat without a focusing lens or linearly with a focusing cylindrical lens. A laser light source can also be used in conjunction with an optical system that produces linear illumination.

The choice of the light source with certain properties can thus be made according to the requirements for the particular application. It must be taken into account that the lighting must be sufficient to identify the markings sufficiently quickly, and that the marking itself has only an area of 6.25 cm ² , can be at different heights on a conveyor belt and with Moving speeds up to 2.5 m / sec.

When the above-mentioned static fixed focus scanner is used, an illumination of about 2 mW / cm ^{2 is} usually sufficient, which can be obtained by a fluorescent lamp.

The second step in the detection during the decoding process is the optical detection of the illuminated area with an electronically operated sensor. The camera or light sensor used in the static fixed focus scanning system of the present embodiment may be an industrial type color CCD TV camera, such as the Panasonic Industrial Co model WV-CD 130 ., which is equipped with a 50 mm TV lens with a 5 mm extension tube, as is available under the brand name NAVITRON TM. Such a camera may be connected to an image capture circuit such as the Data Translation Inc. model DT-2803-60.

The optical image can capture the entire characteristic drawing using an area sensor like that above camera and the image capture circuit hold or alternatively with a linear sensor array leads, which contains a CCD element, the second Dimension when scanning the marking by the Be receive movement of the item (and the marking) becomes. A CCD element suitable for this purpose is the Linear CCD sensor model Thomson-CSF THX 31510 CDZ with 4096 Elements.

For dynamic systems in which the objects carrying the markings are moved on a conveyor belt or the like, it is expedient to provide a long optical path between the markings to be detected and the light sensor. The main reason for this is to reduce the change in the apparent size of the label detected by a spaced sensor. For example, if the length of the optical path is 1.20 m, the image of a label that is only a few centimeters above the conveyor belt has a completely different size than that of an label that is 90 cm above the conveyor belt. However, if an optical path with a length of e.g. B. 6 m is provided, the illustrations of such markings are almost the same size, and it can be obtained a uniformly good resolution of the image in the total area covered. The same applies if an area sensor is used instead of a line sensor. The large resolution is thus achieved with an arrangement as in FIG. 6 by a long optical path.

To focus on different license plates To enable high objects, a height sensor will be compelled. For example, an ultrasonic sensor can be used for this purpose be used, or it is a number of light barriers provided that are interrupted by the object. Correspond An adjustment mechanism with or without back can then be made accordingly guidance (regulation) to be activated to the position of the optical detection elements (e.g. lenses or sensors) continuously adjustable.

Fig. 6 is a schematic view of a setting device for a camera for positioning the light sensor of the camera according to the height of the detected objects. In FIG. 6, a lens 196, a Spulenan control 202, a height sensor 206, and a control therefor are shown. The height sensor 206 may consist of an ultrasound sensor or of a system in which light rays are interrupted by the object. The output of the height sensor 206 is passed to a microprocessor 204 which in turn causes the spool driver 202 to move a spool 200 to which a CCD element 198 or other suitable sensor is attached. A position sensor 208 detects the position of the coil 200 and outputs the corresponding output signal to the microprocessor 204 , whereby a control loop for detecting and adjusting the position of the coil 200 is formed.

The sensor must be capable of being illuminated by the Marking to detect reflected light and it must be a Generate analog signal that reflects the intensity of the reflected Characteristics of the label corresponds to how it passes through the individual pixels of the electro-optical sensor be taken.

A suitable light source can, as mentioned, over a för The tape should be attached to an area that spans the extends the entire width of the conveyor belt, determine with light the quality and intensity. That from the Kenn Drawing reflected light can be captured using a number of Reflectors deflected and only then through the electro optics rule sensor are detected. The purpose of redirecting the It is light to be a compact and therefore rigid system create.

The analog image signal output by the sensor is then filtered. The electrical analog signal is connected used with an analog bandpass filter, the existing to determine whether there is a target in the data field. The analog signal is then processed using a conventional Ana log / digital converter in the image capture circuit or another device converted into a digital signal delt. It is possible to replace the analog bandpass filter to provide a digital filter circuit to the existing being the target by comparing ent speaking digital data with that of the analog / digital con determined data output.

An example of a surface sensor with a CCD element with a number of detectors is the one already mentioned Color CCD TV camera WV-CD 130 from Panasonic. The analogue out This sensor's output signal becomes the image acquisition scarf  tung DT 2803-60 led by Data Translation, the one 6-bit single-color A / D converter for digitizing the data contains. Using a subroutine, the sequential digital output signal of the image capture circuit in the Form of an exact reproduction of the by the optical sensor captured image stored in a memory.

The processing of the optically recorded image is the we Most significant process in the exact recovery of the original initial configuration of the labeling and opti properties (colors) of each hexagon. This Processing includes the steps given below te, the well-known pattern, according to which the labeling was originally encoded, which is used in the Decode information contained in the label.

The first of these steps is to locate the center of the acquisition target. Before the use of the mentioned CCD-TV camera and the image acquisition circuit, as shown in FIG. 10, an initialization program "DTINIT.C" 250 is carried out in order to bring the image acquisition circuit into a defined standby state and to load the color look-up tables followed by the program "DTLIVE.C" 255 to activate the image acquisition circuit. Then the program "DTGRAB.C" causes the image acquisition circuit to digitize the scene into an image memory with 240 lines and 256 columns, the pattern is stored as a byte in 6-bit values. Two auxiliary programs "DTSAVE.C" and "DTLOAD.C" transfer the screens to and from a memory.

When you first register the image of the license plate a conventional analog bandpass filter is used for who the optical properties of the concentric rings identify the acquisition target. This optical egg properties are preferably the colors black and white, because the greatest contrast gives the strongest signal. Around a fixed pattern for the transitions from black to white and find out again after black, etc., it is convenient if a linear scan through the center of the acquisition target regardless of the orientation of the license plate always results in the same frequency response. The capture The best solution is to use concentric rings educated. The sensor output signal is then applied divided and sent through two detection branches. By doing one of these branches becomes the total energy of the output i gnales and in the other the energy with the ring frequency measured. When comparing the two output signals must the energy from the ring detector almost with the energy match the total energy detector when the scan through the center of the acquisition target. In other words, the center of the target is grasped when this almost complete correspondence occurs. Under the file name "FIND.C" a list of the processes for generating a digital bandpass filter and specified for the filtering process ben. In the dynamic scan embodiment is preferably an analog in the first filter step Bandpass filter or an analog sampling bandpass filter used.

It should be noted that the acquisition target location step "FIND.C" 280 of FIG. 10 is designated as optional in FIG. 7 because a hand-held scanner can also be used, with the operator placing the scanner appropriately to ensure correct alignment of the sensor. However, this is understandably much slower than with an automatic sensor, and for high-speed operation, the use of an automatic sensor is unconditional 47658 00070 552 001000280000000200012000285914754700040 0002003943680 00004 47539gbar. If an automatic (not hand-held) sensor is used, the localization of the detection target is a necessary procedural step.

As an alternative to the analog filter described above using a Parks-McClellan algorithm that in the software package "Digital Filter Designs Software for the IBM PC "(Taylor and Stouraitis, Marcel Dekker Inc., New York N.Y. 1987) contains a digital bandpass filter be generated.

In the present embodiment, a one-dimensional digital bandpass filter was used to filter the normalized digital bit stream using the following filter subroutines. The band to be filtered out is the expected ring frequency. The one-dimensional digital bandpass filter is intended for a sampling rate of 400 pixels per inch (2.54 cm) and a length of 125 pixels (7.9375 mm or 0.3125 inches) and designed for the dimensions of the printed acquisition target rings, as in As shown in FIG. 3. The frequency is therefore 300/16 line pairs per inch (2.54 cm), which corresponds to a normalized frequency (for which 400 line pairs per inch or 2.54 cm = 1) of 300/16 × 400 or 0.046875. A filter with a passband is selected which extends up to 5% below this frequency and 15% above this frequency, since distortions in the marking typically result in a reduction in the image and thus an increased frequency. Exclusion zones of 15% below the frequency of 0 and of 25% above the ring frequency of 0.5 (Nyquist limit) were provided. The values for the filters are stored in the file "IMPULSE.LUT" 275 ( Fig. 10) for later use, the first 62 coefficients being omitted because the filter is symmetrical. A corresponding flow chart is shown in FIG. 8.

By sampling the bandpass filter at output intervals, that correspond to the measured horizontal magnification, a filter with a length of 25 pixels is generated.

If, for example, the horizontal enlargement of Fig is 80 pixels per inch or 2.54 cm, each fifth sample of the filter used (400/80 = 5 image Points). A linear is used for non-integer steps Interpolation of adjacent filter samples performed.

Furthermore, a second, two-dimensional filter is from 25 × 25 pixels are provided. The samples for this two-dimensional filters are based on the Euclidean Ab stood of every point from the center of the filter, the for a suitable horizontal and vertical magnification was discontinued. For non-integer sampling intervals linear interpolation is performed again.

The output signal of the mentioned one-dimensional filter is squared and using a recursive low-pass filter first order smoothed out what an exponential out cut from the history. If the output signal of the Smoothing filter exceeds a predetermined threshold an additional two-dimensional filter step is provided, to check the existence of the target and to determine its location exactly. In the first part of the two dimes sional filtering process is a reduced filter size  of 10 × 10 pixels used to calculate it easier. This filter scans a rectangular area the location determined by the one-dimensional filter from. If the maximum of the two-dimensional correlation is one exceeds the predetermined threshold, it will eventually become full constant two-dimensional filtering with the full 25 × 25 Pixel filter around a small square window the maximum applied. If the best result of this fil in turn exceeds a predetermined threshold the center point. If one of the thresholds is not is exceeded, the program "builds" the smoothing filter partially and returns to one-dimensional scanning return When the one-dimensional scan is completed, without the existence of a detection target the program ends with an error return Leap.

The intensities of the reflected light recorded by the optical sensor used can change due to changes in the lighting, the printing density, the reflectivity of the paper, the sensitivity of the camera and other factors, including deterioration in the marking, for example due to folding, warping, etc. . As an optional (and preferably applied) step, the reflected light detected by the sensor and the corresponding signal passed to the memory can be normalized in a conventional manner. Using known techniques, the normalization program "NORM.C" 270 shown in FIG. 10 is used in the analysis of the intensity level of the light reflected from the label as recorded in pixel blocks in the scanner to determine the minimum and the determine the maximum intensity of the light reflected by the data field. The digital output signal of the above-mentioned combination of scanner and image acquisition circuit is reloaded from the memory to the computer in order to be further treated by means of the normalization program.

Using the equation y = mx + b, where the in x minimum intensity used a value of y = 0 and the maximum intensity used in x has a value of y = 63 results, the detected intensities of the reflected Light for each pixel set so that it is faint zeste black and the whitest white in the saved Figure set as standard and the others Shades of black, white and gray are based on these standards be drawn. The normalization process is therefore easier the processing of the captured image. The normalization is executed using a program "NORM.C". It can of course also other, more complex normalization procedures are applied.

For the following calculations the saved re Handing the label rescaled to an illustration available with the same horizontal and vertical sizes. Again, this is an optional step, but one fast and accurate retrieval of the coded information tions relieved. The scaling is carried out in the picture has an even horizontal and vertical Resolution of the scan of, for example, 150 pixels per To get inches (2.54 cm), as when running with the static fixed focus scanner is used.

The scaling or re-scaling is done by calculation the partial row and column addresses of samples 1/150 inch based on the well known horizontal and ver  tical enlargement. Every point of the new evenly scaled figure is then from a corresponding sentence derived from points from the figure that are in memory is included. For partial addresses it becomes approximate Determination of the value of the points a bilinear interpola tion used. The scaling places the center of the Marking in the memory in a known position. The scaled figure will eventually be for later use saved in the search step. All of the following assume that the scaled image of the indicator centered on a known position in the grid, it it should be noted, however, that the orientation of the characteristics drawing is not yet displayed, which relates to the Sen sors can still be tilted. The scaling will executed using a subroutine.

The next operations are collectively referred to as "two-dimensional clock recovery". These steps are performed by a program and subroutines titled "CLOCK.C" 290 ( Fig. 10). This process is carried out two-dimensionally on the back-scaled image in order to precisely determine the position of each hexagon in the original data field. The purpose of clock recovery is to determine the scan positions and correct the effects caused by warping, curling or tilting of the label since the label, for example, need not be completely flat. This is an important part of the process, its application is not limited to hexagonally coded markings, it can also be applied to other markings with a regular two-dimensional grid of, for example, squares, triangles, etc.

The one-dimensional clock recovery is at the signal processing a generally known process. A two-way Dimensional clock recovery is an extension of this Process that can be carried out by any specialist. It it should be noted that the term "clock recovery" can be a little confusing for non-professionals since they don't timing concerns.

The first step in executing the clock recovery can be determined using known, non-linear assignment gears are carried out to signal components with a To generate clock frequency in the digitized picture-off output signal of the optical sensor and the image acquisition circuit are not available. The purpose of the non-linear It is assignment that is preferably normalized and back scaled figure that at this point in the process is to take and into a two-dimensional not Convey linear illustration or map that shows the transitions between neighboring contrasting hexagons lifts. This is done in the present embodiment through an assignment using the standard deviations, the is called a mapping of the standard deviations. This before gang can also be filtered by using a dif referencing core, of which several types are known, such as LaPlace or Sobel cores, and a determination of the absolute worth or squaring the results the. Appropriate procedures are in "Digital Image Pro cessing "by Rafael G. Gonzalez and Paul Wintz, Addison Wesley described in 1977.

In the ordinary mapping of standard deviations is the image with undifferentiated cell borders in the Memory saved. It then becomes a picture of Stan  dard deviations generated to the edges of adjacent kon tracing hexagons by determining the standard softening of 3 × 3 groups of pixels (this is something other than the groups of 3 × 3 cells) to determine the standard deviations of the intensities of the pixels to be determined. The standard deviation calculations are used to define the areas of pixels with a specific color (the smallest standard deviation) det that the inside of a hexagon or the transition between represent two hexagons of the same color, in the opposite set to the groups of pixels with larger standard deviations, the transitions from a hexagon to a certain color to an adjacent hexagon with a represent contrasting color. Because side by side Hexagons often have the same color, give a picture The standard deviations do not complete every hexagon dig on. Missing borders or borders between hexagons arise from the fact that the process of mapping using the standard deviations transitions between six corners of the same color. The clock Recovery is therefore missing to create this Transitions provided - The decoding process described can be applied to any of the above described forms of labeling are used. The Coding units for the different geometries can can be easily adjusted because the optically coded polygon cells are arranged so that the geometric mean points of neighboring polygon cells on the corners of a given two-dimensional grid.

If the labels with optical sensors called the The respective geometry is read out or shape of the individual coding units or polygon cells  not determined by the optical sensor. The sensor picks up instead only the labeling with a known one Number of samples per inch or centimeter on and records the intensities of the optical Properties of the respective scanning area reflected Light on. These values are used in later stored in a memory. In other words, the electro-optical sensor the average light intensity Scanning area for scanning area over the entire marking on, regardless of whether anything at all the marking is printed out. This is under the spit Saving the image with not differentiated in the memory Transitions from cell to cell are meant. For this reason the decoding process easily on labels of different Adjust configurations as long as the centers of the Po lygon coding units a predetermined distance and one predetermined direction in a two-dimensional grid to have.

It has been found that modifying the hexagonal cell-based system, as in the case of polygons that are only substantially hexagonal in shape, as shown in FIG. 15, results in little degradation in the performance of the Systems results. The use of polygons with poor packing properties or grids or partially or not at all connected figures results in a bad re, but nevertheless sufficient performance for some applications. At some point, however, the performance of the system naturally falls due to optically no longer resolvable high-frequency components of lower order polygonal coding cells, with poor cell packing and with grids with large gaps between the polygons, to an unacceptable, too low capacity for the storage and recovery of the information.

The suitability of the system depends on the quality of the signal picked up by the electro-optical sensor. By changing the detection system, for example by Increase the number of samples per unit area The surface of the marking can be opened by the sensor recorded signal improved and the possibilities of Infor Mation storage and recovery in particular of partial and non-contiguous labeling confi gurations can be improved.

The system therefore enables a wide range in terms of the subject of the labeling, the facilities for Recording the optical signal and processing the optical signal. Polygon cells from either right irregular or irregular or irregular shape as Coding units are used. As long as the distances and Directions of the centers of the polygons with respect to neighboring barter polygons are known, the polygons on ir grid, and they can be complete, partial wise or not related at all.

The nonlinear mapping techniques, especially the be written illustration of the standard deviations, facilitate the reconstruction of the missing transitions or borders between the polygon cells with the same optical properties ten using the procedure described below. On the same way can the missing transitions between polygons and spaces with the same optical Properties are found when the polygons are partial or not related at all. That is through the im Fast Fourier Transform, Fil ization and inverse fast Fourier transformation achieved.  

An optional technology reduces the generation the mapping of standard deviations required Calculations. Usually they are used to calculate the sum eight of the nine pixels in each of the 3 × 3 pixel blocks Additions required. This can be cut in half be when each pixel of the figure by the sum from himself and the one immediately to the left and right of it lying pixels is replaced. This requires two ad editions per pixel. The same process is then carried out on the new figure executed, except that the sum for the pixels directly above and below te is calculated. This adds two more and therefore a total of four additions are required. At the end of The process is every pixel by the sum of itself and replaced its eight immediate neighbors.

The mapping of the standard deviation becomes the generation an illustration of the hexagons that match the original data field corresponds, but no transitions between hexagons contains the same color, preferred.

The following subroutine, called "fenestration", is again optionally provided. The fenestration or detail representation is used to reduce the intensity of border lines used that do not correspond to hexagon outlines. This Limits occur in two places: the acquisition target rings and in the illustration of the uncontrolled environment of the Labelling. To reduce the intensity of this area a weighting function is used. Like a fen before a fast Fourier transform is known to the person skilled in the art.  

Under the control of a commercially available program is then shown on the (optionally windowed) image of the Standard deviations a two-dimensional fast Fourier Transformation of digital values carried out. This leads a computer with the figure in the previous step was generated using a fast Fourier transform a two-dimensional representation of the distances, directions and intensities of the transitions of contrasting hexagons to get. Put simply, the fast Fourier transform is mation a measure to determine the distances, dir edges and intensities of the edges between hexagons, insofar as known. The regular distances and directions of the limits tongues of the hexagons result in points in the transformation concept rich with a high energy level. The brightest point is is accordingly in the transformation level at 0.0 the DC component of the figure. The the center surrounding six points represent the distances, directions and Intensities of the edges between the hexagons.

By performing a fast Fourier transform the digital data, that of the non-linear, he can also be a two dimensional representation of the distances, directions and intensities compute the interfaces of the contrasting polygons as in the previous mapping step over which standard deviations were identified. The Ab The states and directions of the polygon boundaries then result certain points of high energy in the transformation area. The number of high energy points that make up the zero point Surrounding 0.0 of the transformation plane depends on the geome trie of the respective polygon cell that is used to build the Labeling is used. In the case of hexagons, these represent the points surrounding the center point the distances, directions  and intensities of the edges between polygons or between Polygons and spaces when marking out only partially or not connected elements is building.

Since the mapping is valued real (and not complex), is the transformation area around the origin point-symmetry trisch. It only takes half of the transformation area to be calculated, which corresponds to the computing time accordingly reduced. This also reduces the effort in the fol image filtering and the inverse fast fou reduced rier transformation. That in connection with the static fixed focus system used program with a fast Fourier transform is, for example Subroutine R2DFFT of the 87FFT-2 package from Microway Inc., Kingston, Massachusetts.

To reconstruct the full outline of all hexagons in the image area using the transformed di filtering data. Sol eliminated points in the transformation area, which the ge desired distances and directions of limitations of the Hexagons in the mapping step with the standard deviation received, do not correspond. There is in Transformation area due to the hexagonal honeycomb structure the marking six distinctive points. In transformation only three points are actually identified graces because the image is point symmetrical around the origin and the second three points are derived from the first three can be. Filtering is preferably done in three Steps taken to make such transitions from across the To eliminate standard deviations obtained mapping that are too far apart or too close to each other and / or point in the wrong direction.  

First there is a high pass filtering in which all points within a certain circle around the origin of the Transformation area can be set to zero, however without the six striking points that appear in the graphic trans formation area at a distance outside the origin are arranged in the shape of a hexagon. These points correspond to distances that are greater than the distances in the hexagons and thus the information about the missing contain the transitions in the image of the marking. At feh to recover lingering transitions in the image of the marking, it is necessary to provide information about the missing To eliminate transitions in the Fourier transform range.

Next, all the points are outside of a certain Radius around the six striking points in the transformation area set to zero. These points correspond to disturbing Transitions that are too close to each other. Along with the first surgery is through this surgery a ring of remaining points is formed. The production this ring is performing a spatial bandpass filtering equivalent. The inner and outer radius this annulus is determined by means of the expected distance Hexagon outline determined. Since it can be expected that in the example described the "diameter" of the hexagons is equal to 5 pixels and there for a transformation length of 256 pixels, the tips of the hexagons in the trans formation range equal to 256/5 = 51.2 pixels from the mean point should be removed, corresponds to a ring with a inner radius of 45 pixels and an outer radius of 80 pixels the hexagon diameters used by 3.2 to 5.69 pixels. A filter is used that also allows higher frequencies, because deformations of the characteristic drawing like warping and tilting downsizing  of the picture.

After performing this spatial bandpass filtering remains a circular ring with six striking points, one of which each to the center (zero point) of the transformation area ches has an equal angular distance. For complete Eliminate unwanted information in the transformation directional filtering is still carried out in the area. Everyone Point with too large an angular distance from the striking Points in the transformation area are set to zero. The has the consequence that all edges in the image area are removed those that are not in one of the three directions, the occur in the hexagon honeycomb pattern.

It is required to perform this directional filtering Lich to find the most striking point, which according to the space union bandpass filtering is left. It is believed, that this point is one of the six striking points in the trans formation area is the ent of a hexagon speak. Five also occur in the transformation area other prominent points with the same distance from the center point and the same angular distance of multiples of 60 Degrees on. Therefore, all other points with an angle were more than 10 degrees from any of these points eliminated. There remain six wedge-shaped pieces of the Ring. With this directional filtering, information about wrong distances or directions in the image area does. Eliminating this information about incorrect ab stands enables the recovery of the complete Outline of each hexagon in the image area.

The filtering mentioned is carried out using subroutines leads.  

The filtering process described above for coherent hexagons must be changed if other two Dimensional grids can be specified for labeling. However, there are other configurations to consider only minor modifications required.

After the individual polygon cells have been defined the angular distances of the boundaries and the number of Be ten and their lengths are known. Next it is needed the relationships between neighboring polygons determine, for example, whether they are related, partially related or not related. That is also geo metric grid, according to which the centers of the Polygons are arranged. Because the geometry of the label the appropriate filter can be used to stop filter the energy points in the transformation area easily be created so that only the brightest points correspond the appropriate distances and directions of the polygon boundaries through the subroutine for inverse fast Fourier Transformation.

As for the filters generated, it is necessary a suitably dimensioned spatial bandpass filter the basis of the given distances of the coded polygon to generate cells. Then it is convenient to use a directional filter to filter out the energy points that are not the markan test points are the axes of the given two correspond to the dimensional grid. Because of who the information about incorrect distances or directions of the Cells in the image area and, if applicable, the gaps eliminated. By eliminating this misinformation can have a full grid of resources in the image area points of the polygon cells using the inverse fast  Fourier transform can be obtained.

To return to the image area and thereby the image the outline of the adjacent hexagons of the data Restoring the field becomes a two-dimensional in verse fast Fourier transformation on the filtered Data of the transformation area executed. This inverse Fourier transformation takes place by means of a corresponding one Standard subroutine included in Microwave's 87FFT-2 package Inc. is included. After executing the inverse Fourier Transformation are the outlines of each hexagon in the Image area restored. Have in this new picture the centers of the hexagons are of great value. That did neuter values for the points at the midpoints of the Hexagons depend on how many edges are around position. More edges create a higher energy level at the permitted frequencies and therefore points with larger values. Fewer edges result in points with smaller ones Values. The values of the points are a good measure of the value the probability of information during clock recovery at any given point.

The image of the hexagons has now been recovered, it is however, it is still necessary to determine their orientation.

The hexagonal honeycomb pattern has three axes with an angular spacing of 60 degrees each. The direction of these axes is determined by the brightest points in the transformation area after spatial bandpass filtering. It is then possible to find out which of these three axes is the main axis. This step is optionally carried out. If the step is not carried out, the label is to be decoded three times using each of the three axes, only one axis giving information with a meaning. The main axis is arbitrarily chosen as the axis of the hexagons, which runs parallel to two sides of the marking, as was described in connection with FIG. 2.

If the limits of the square marking on the Based on knowledge of the major axis, then determined most of the energy is in the restored pattern of Hexagon outlines within these limits of the square.

To determine the main axis, it is first assumed that each of the three axes can be the main axis. For each of the Axes will then tentatively become the outline of the resulting square marking determines and ascertained how much of the total clock recovery energy from the digi tal energy data from the subroutine for the inverse Fourier transform can be obtained within this Square. The right axis is then that which yields most of the energy. The angle of this axis is then for initialization and other searches are saved. It it is not yet known whether the recorded angle is in the right direction or wrong by 180 degrees is. It is understandable that not all three marking areas need to be determined since it does not require energy in the areas to determine all three squares are common.

A program titled "SEARCH.C" 300 ( Fig. 10) combines the transformed and restored information about the hexagon centers with the stored intensity levels of the original image to determine the value for the gray tone of each hexagon. The search is performed so that the probability of "getting lost" during the search is minimal. The end result is a matrix of grayscale values for each hexagon of the data field. In the first part of this program, four important information fields are set up. The field CVAL (clock value) stores a value for the quality of the recovered clock signal for each hexagon, while the field GVAL contains the gray tone values (0-63) in the center of each hexagon. The row and column positions of the center points of each hexagon are stored in the remaining fields IVAL and JVAL.

From the determined main axis angle and the known Ab the hexagons (5 pixels) are the expected horizontal and vertical distances from the center of a Hexagons to the centers of the surrounding hexagons calculates.

The SEARCH.C program works on these calculations the clock recovery signal coming from a memory and the scaled-down image of the marking, which is also obtained from a memory. The main The purpose of the initialization subroutine is to merge and summarizing the information from these two sources and the Generate a data matrix that shows the grayscale values for each Contains hexagon.

The search initialization step is around a square the center of the label with a side length of be about a third of the side length of the label borders. Within this area is a good starting point the point with the greatest value in the recovered taktsi  signal field. Then the location of this starting point is relative to The center of the marking is determined. This starting point is a point where the clock signal is strong and clear, and also a point that is the center of the marking is relatively close. A clear signal is cheap to ensure that the search with a legal six corner center begins, and the location of this point close to The center of the label ensures that its location can be determined without distortion or jamming exert greater influence. A measure of quality of a point in the clock recovery pattern is that Value for the point minus the values for the eight surrounding ones Points. The rectangle coordinates of the starting point are in Polar coordinates converted, the polar coordinates regarding of the previously determined main axis angle and that The result is converted back into the rectangular shape. This Coordinates are made according to the expected line spacing (4, 5 pixels) and column spacing (5 pixels) scaled, to go to the insertion positions in the hexagon matrix come. The clock quality, shades of gray and positions that the Start hexagon match, then match in the the fields CVAL, GVAL, IVAL and JVAL.

The following main search loop leads to localization the centers of the remaining hexagons. The bow will end when the expected number of hexagons lo has been calibrated. The order of the search for the The center of the hexagons is extremely important. The increased reliability of the decoding process in the event of faults the labeling results from the particular used Search technique, which is explained below.

Each iteration of the search loop begins with the new one Call the location of the clock recovery point with the  highest value, whose neighbors are not yet at their highest Value have been searched. From this known point the search for a hexagon for each of the six directions continues. The search pattern is thus along a path built, the recovered from better to worse Beat quality leads. If there is a bad area in like the clock obtained, for example in the center of the Marking or in a disturbed area, the runs Search algorithm around it and not right through it. By bypassing the bad areas and picking them up last, the likelihood of getting lost on considerably reduced the grid. Because getting lost is the same is as bad as reading a wrong shade of gray this property of the search algorithm is extraordinary significant.

To find the neighbors of the best clock values with the If a main loop is found, a subroutine is used det. The subroutine loops six times, once for every adjacent hexagon of the six under consideration corner. First, the location of the neighbor is calculated. If this neighbor is outside the limit of the flag iteration, the iteration of the loop is ended. If not, the neighbor is checked to see if he is already out another direction was detected. If the neighbor be has already been recorded, the iteration of the Loop also ended because the previous search algorithm runs rated higher than later ones. If the neighbor has passed these tests, the expected position of the Center of the neighbor in the clock recovery pattern calculated. At this point a gradient search for the Clock signal with the largest value executed. The again found position surrounding eight pixels are on it  searches whether a higher clock value is found. If yes, the eight neighbors will be the best neighboring point checked whether an even better value is not found. The This gradient search gives a degree of adaptation that is un is necessary if distorted and tilted characteristics drawings can be read. The subroutine then goes to nearest neighbor over or jump back if all neighbors have already been checked.

As described above, the reconstructed git contains ter as the result of the data transformation about the geometric centers of the polygon cells. The This grid points in areas where originally more contrasting interfaces were available, more energy. The center points lie on the given two-dimensional len grid with a given number equal or not equally spaced axes. The information about the spatial relationships of the axes of the given grid can be used to determine the orientation of the main axis be attracted.

However, it should be noted that the algorithm changed in this way can be that the actual geometry of the two-dimensional grid is determined and that the filter scheme is determined from this determination so that the so-called main axis of labeling, that is, the axis of the two-dimensional grid, the pa parallel to two sides of a square marking, and the necessary coordinates for the search subroutine be generated.

Whether the geometry of the marking by such a choice step as determined above or whether simply by appropriate modifications of the two-dimensional clock re-  extraction process is entered into the decoding process, can always use the different types of labels can be easily processed. The number of axes on which the centers of the individual polygon cells lie, and their respective orientation can be used in determining the Main axis are used. The main axis of the previous one The two-dimensional grid can be determined without performing the above empirical analysis process becomes.

In the hexagonal arrangement according to the preferred embodiment form of the information can be obtained from the determination of the Main axis and the known distances between the polygons used, the expected horizontal and vertical Distances from the center of a polygon to the center of the surrounding polygons. According to such calculations and after making the necessary changes the search subroutine will do the search including the ini initialization step and the main search loop for the respective labeling configuration executed.

After the subroutine has been completed, the position of the current center is marked so that it is not searched again. This is to delete this position as a candidate for a neighbor search. With each loop iteration, 0 to 6 new candidates are added and one candidate is deleted. An effective implementation can use a data structure that includes the candidates in order of size when the insert and delete operations are performed. One of the possible structures is the so-called priority queue (cf. "The Design and Analysis of Computer Algorithms" by Aho, Hopcroft and Ullman, Addison Wesley 1974). A linear search algorithm is known to require n ² operations, while an efficiently executed priority queue with a symmetrical tree or heap structure requires nlogn operations. A search algorithm based on memory sorting with n operations can also be used if the recovered clock values are scaled and reduced to a small range of integers.

After the main search loop is completed, the Location of all centers of all hexagons determined and it are the shades of gray of the centers of all hexagons that saved, fully used. The next Step is a threshold determination of the digitized Grayscale values in the range from 0 to 63 for the discrete values for example black, gray and white. this happens by building a histogram for the intensity values of the Hexagon centers of the marking. Cutting level can determined by detecting dips in the histogram become.

After a threshold value determination of the discrete levels two faults are still present. For one, can the array must be shifted from the center. The can Case if the initial search step is the location of the best quality clock signals relative to the center of the Identification does not exactly indicate. The second way be is that the entire label is upside down was read because the major axis is related to the 180 degree Angle is not clear.

A subroutine determines whether the label from the Center point is shifted. If the labeling is correct  the coordinates of the middle line run through the center of the label. To realize, Lines become whether there is a vertical positioning error checked over the adopted center line whether it form a line closer to the label center passes by. If a line above or below is closer to The center lies as the adopted center line a corresponding shift up or down. If the left alignment of the short lines was wrong is, this is done by moving the short lines around corrected one position to the right.

Horizontal positioning errors and a reading "upside down" are checked using information which is contained on the label as so-called coarse grid information. The information is divided into groups of 3 × 3 cells, as already described. Since the labeling in the given example represents a grid of 33 rows and 30 columns, these groups form an 11 × 10 grid. The lower middle hexagon of each complete group of 3 × 3 cells has a certain property that is created during coding. As described in connection with FIG. 4, this hexagon has a fixed transition on each side. For example, if the lower middle hexagon is black, the lower left and right hexagons are either gray or white. A subroutine perceives these transitional properties to remove the two possible faults. For this purpose, an array is first created in which each element of the array indicates whether there is a transition between two horizontally adjacent hexagons. Then the array is checked for each of the nine hypothetical sections of the coarse grid arranged as a 3 × 3 pattern for the expected section of 0. One of these cuts will show a better match between the actual and expected transitions, and the location of this cut will be determined. The same hypothesis is then tested on the assumption that the label has been read upside down.

If the label is simply turned over, that is if the top lines with the bottom lines and the higher ones Columns swapped with the lower columns is that The result of the pattern formation turned upside down. It must be the correction of an upside down license plate, however an important transformation to be carried out. During the Reading the short lines (length = 29) to the left directed; therefore, in the event of a reversal, the indicator these lines are aligned to the right. This The result is that the result of the adopted Cuts is different than a simple reversal. The best result of the cutting tests will be better than any previous test if the marking is actually has been read upside down.

After determining if the label is upside down was read and whether the cuts in the absolutely rich current position, the matrix of the marking be decoded. With the correct definition of the figure and the editing is finished, and data decoding processes begin.

A program "RD.LABEL.C" 182 ( FIG. 9) reads out the file generated by the search program and generates a bit sequence file with, in the present embodiment, 1292 bits. For this purpose, a subroutine "CELL DEC.C" 183 is used to hide hexes that cannot be used and for the decoding process, which represents a reversal of the coding process.

The first step in decoding is generation a bit string of information through the six corner are shown, a mapping process of Hexagons are used on bit patterns, which is a reversal the process of mapping the bits onto the hexagons that was used for the encoding.

The program then converts the bit sequence into a bit sequence for the High priority information and a bit string for the In formations of low priority or as many bit sequences shared as used during coding the label were detected. It is then necessary to use the Error coding techniques used when coding the identifier error correction on each bit follow to execute. For example, if a Reed-Solomon code error correction on the bitfol output generated by the search program gnal, which has the same format as above written coding input file. An error correction can with the following operations (see "Theory and Practice of Error Control Codes ", see above):

• 1. Calculate syndromes
• 2. Calculate the error localization polynomial using the Berlekamp-Massey Algorithm;
• 3. Calculate the location of the errors using a Chien search; and
• 4. Calculate the size of the errors using the Forney algorithm.

The last step is only carried out if a correctable number of errors was found in steps 2 and 3 . The number of errors found is also calculated. If an uncorrectable number of errors is found or there is an error within the padding (see above), a flag is set. The error procedure used has the designation "ERRDEC.C" 184 ( FIG. 9).

In order to output the decoded information on a computer terminal, the program "TEXTOUT.C" 185 ( FIG. 9) can be used.

Claims (13)

1. Optically readable identification for storing coded information, with a data field which is provided with a number of polygons, each of which has one of at least two different optical properties, and which lie with their centers on a two-dimensional grid, characterized in that the polygons regular or irregular polygons of any order and arranged on a hexagonal grid. 2. Optically readable identification for storing coded Information, with a data field that is based on a two-dimensional Lattice arranged polygons is provided each one of at least two different optical properties exhibit, characterized, that the angles between the sides of the data field are not agree with the angles of the grid axes, so that a the grid axes by their angular position to the sides of the Data field in front of the other grid axes as the main axis is excellent. 3. Labeling according to claim 1 or 2, characterized in that the polygons are essentially in the form of hexagons have and arranged according to a honeycomb pattern are.   4. Labeling according to one of claims 1 to 3, characterized characterized in that the polygons are partially contiguous or are not contiguous and that between the poly gonen spaces with one of the two op table properties are available. 5. Labeling according to one of claims 1 to 3, characterized characterized in that the polygons are arranged contiguously are. 6. Labeling according to one of the preceding claims, characterized in that the grid is three axes with an angular distance of 60 degrees each. 7. Labeling according to one of the preceding claims, characterized in that the optical properties of the Colors are black, white and gray. 8. Labeling according to one of the preceding claims, characterized by a number of concentric rings, that occupy an area on the label that the Area is separated on which the information is housed where each concentric ring is one of at least two different optical properties in alternating order having. 9. Labeling according to claim 8, characterized in that the concentric rings in the middle of the marking are arranged. 10. Apparatus for detecting and decoding the optical identification according to one or more of claims 1 to 9, comprising:

• a) a device for illuminating an area which is intended for the marking to pass through;
• b) a device for optically imaging the illuminated area and generating electrical analog signals which correspond to the intensities of the light reflected by the polygons, which falls on the pixels of the imaging device;
• c) a device for converting the electrical analog signals into a digital bit sequence which corresponds to the intensities of the light received by the pixels of the imaging device;
• d) a device for storing the digital bit sequence for the subsequent decoding of the identifier; and
• e) a device for decoding the digital bit sequence, this decoding device generating an electrical output signal which represents the coded information,

a device for detecting the center of the marking is provided and the decoding device by the digital decoding the orientation of the data field certainly. 11. A method for producing the optically readable marking according to one of claims 1 to 9, characterized by the method steps

• a) mapping one of the at least two different optical properties to each polygon to produce a number of polygons with different optical properties;
• b) encoding the information by ordering the polygons in a particular sequence; and
• c) printing each polygon with the associated optical property.

12. Use of the label according to one of claims 1 to 9, in a method for storing and retrieving data, characterized by the method steps

• a) printing the marking with the polygons which are coded according to a specific coding process;
• b) illuminating the label;
• c) optically detecting the light reflected by the polygons with an electro-optical sensor;
• d) generating electrical analog signals which correspond to the intensities of the reflected light as it is detected by the individual pixels of the sensor;
• e) converting the electrical analog signal into a sequence of digital signals;
• f) storing the digital signals in a memory connected to a computer to form a representation of the digital signals;
• g) decoding the representation of the digital signals in order to recover the characteristics of the intensities, positions and orientations of the respective optical properties of the polygons; and
• h) generating a digital bit sequence, which the computer outputs and which reproduces the decoded information represented by the polygons.