JPH07192133A - Circular pattern discriminating method - Google Patents

Abstract

PURPOSE:To realize an excellent dicrimination ratio while shortening the discriminating time of a circular pattern. CONSTITUTION:A circular pattern is read, the density of each picture element is stored as a digital signal, a part showing the feature of the circular pattern excellently from the stored information is segmented in a circular or ring shape with the center coordinate of the circular pattern as the center, two-dimensional basic data is obtained, a processing which is necessary for this basic data is performed and data array for discrimination of the same size as a binary template is obtained. A binary conversion is performed for this data array for discrimination by using the value of a template : data of the number 1, a comparison pattern is obtained, the matching calculation of the comparison pattern and the template is performed from 0 to 2 for the angular coordinate of the circular pattern, the similarity or the non-similarity of the comparison pattern and the template is detected. Based on the detected similarity or non-similarity, whether the circular pattern includes the pattern according to the template or not is decided.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a "circular pattern identification method". INDUSTRIAL APPLICABILITY The present invention can be used, for example, for discriminating coins and discriminating yen bills for games in vending machines and the like.

[0002]

2. Description of the Related Art Conventionally known circular pattern identification methods related to the identification of coins and gaming yen bills include a feature extraction method and a pattern matching method. , In the case of an identification target that has individual differences in the surface state due to the length of the distribution period and the usage period, the contrast of the captured pattern data differs depending on the surface state of each coin / yen bill, so even if simple pattern matching is performed , A good identification rate cannot always be obtained.

As a measure for enhancing the discriminating power, it is conceivable that the data of each pixel in the reading target area is multi-valued so that the read data faithfully corresponds to the identifying target. This causes an unfavorable problem of extending the processing time.

[0004]

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned circumstances, and a novel circular pattern capable of shortening the data processing time and realizing a good identification rate. The purpose is to provide an identification method.

[0005]

The circular pattern identifying method of the present invention is a method for identifying a "circular pattern" such as a pattern of coins or game yen bills.

As shown in FIG. 1, the method of claim 1 has an input step, a cutting step, a binary conversion step, a calculation step, and an identification step.

The "input step" is a step of reading a circular pattern to be identified and storing the density of each pixel as a digital signal. The "cutting step" cuts out "a portion that shows the characteristics well" of the circular pattern into a circular or ring shape with the center coordinates of the circular pattern as the center, from the stored information, to obtain "two-dimensional basic data". In this step, necessary processing is performed on this basic data to obtain an "identification data array" of the same size as the binary template.

The "binary conversion step" is a step of binary-converting the identification data array using the "value: 1 number of data" in the template to obtain a "comparison pattern".
The “calculation step” performs matching calculation between the comparison pattern and the template “from 0 to 2π with respect to the angular coordinates of the circular pattern (with the central image as the origin)” to determine whether the similarity between the comparison pattern and the template or not. This is a step of detecting the degree of similarity. The "identification step" is a step of determining "whether or not the circular pattern includes a pattern according to the template" based on the detected similarity or dissimilarity.

In the cutting process, the "center coordinates of the circular pattern", which is a reference for cutting, is necessary. In this case, in the case where the reading position of coins or game yen bills is "mechanically fixed", the central coordinates can be uniquely determined in advance. When the reading position is not fixed and the center coordinates of the circular pattern fluctuate with each reading, "The center coordinates of the circular pattern are detected in the cutting process from the information stored in the input process. It is sufficient to do so (Claim 2).

The size of the binary template is "2
The size may be the same as that of the "dimensional basic data", and in that case, "the basic data itself is used as an identification data array" (claim 3). Further, the processing for obtaining the identification data array from the two-dimensional basic data may include differential absolute value sum creation processing and / or data compression processing (claim 4).

Further, the binary template is not limited to one type, a plurality of types are prepared, and a plurality of types of identification data arrays are cut out in one-to-one correspondence with each template, and each identification data array is extracted. It is also possible to detect the degree of similarity or the degree of dissimilarity between the comparison pattern obtained in step 1 and the corresponding template (claim 5).

[0012]

[Function] As represented by the pattern of coins and yen bills for games,
Generally, a circular pattern has a "portion that best represents the characteristics of the circular pattern" in a part thereof. Therefore, in identifying a circular pattern, it is not always necessary to make the entire pattern into a template for pattern matching, and it is sufficient to perform pattern matching only for "characteristic portions".

However, in the circular pattern, it is not clear in which direction from the center of the read pattern information the "characteristic portion" exists due to the rotational symmetry of the pattern shape. Is a “circular or ring-shaped” that includes a characteristic part depending on the distance from the center coordinates of the circular pattern.
The portion of is cut out and this portion is the target of collation with the template. Therefore, the template itself is also created based on the "characteristic part".

Another feature of the present invention is that the template is a binary template which is binarized by "0 and 1", and the "comparison pattern" to be compared with the template is "value: 1" in the template. It is formed by performing binary conversion on the above-mentioned identification data array using "number data". By doing so, it is possible to effectively prevent an identification error based on the “individual difference in contrast” of the circular pattern itself.

The binary conversion for obtaining the "comparison pattern" from the "identification data array" is one of the features of the present invention as described above.

The binary template uses "ideal read data" of the circular pattern to be identified as basic data, and binarizes "data obtained by the same method as obtaining the basic data for identification" from this. It is created by doing.

For simplification of description, taking the case where "basic data itself is used as an identification data array" as in the case of the invention of claim 3, in this case, 2
The value template is the binarized basic data cut out from the ideally read circular pattern.

Here, the "basic data cut out from the ideally read circular pattern" consists of data for N pixels, the data for each pixel is density data, and the density level is for 8 bits, that is, It is assumed that there are 256 levels.

The pixel address is the address number: I (1 ≦
When I ≦ N) and the density level is J (0 ≦ J ≦ 256), the identification data array is represented as a set of data (I, J). The N pieces of data thus obtained are made into a histogram by density level: J. That is, imagewise, the horizontal axis represents density levels 0 to 25.
6 is taken and the number of pixels (frequency) belonging to each density level is taken on the vertical axis.

In the histogram data, the frequency: n is a function of the density level: J and can be expressed as n (J).

Therefore, a specific density level: J ₁ is selected, and the density level larger than this density level: J _{1 is} j.
Number of pixels having (J ₁ ≦ j ≦ 256): sum of n (j):
Σn (j) is set to a predetermined ratio with respect to the total number of pixels: N, for example, 30% (a binary template is experimentally set so as to best represent a circular pattern). The density data of the pixel having the density level thus selected is 1, and the density data of the pixel having the other density level is 0.
And

In this way, a binary template having binary values of 0 and 1 is obtained as the density data. This 2
The valuation method is one of the well-known "P-tile method (industrial image processing: written by Masato Ejiri)".

Similarly, the circular pattern is read for identification,
The cut out basic data (used as an identification data array) also includes data for N pixels, and the data of each pixel is divided into 256 levels by density data.
Therefore, the address of the pixel in this case is the address number: L
(1 ≦ L ≦ N) and the density level is K (0 ≦ K ≦ 25
6), the identification data array is the data (L, K).
Will be represented as a set of.

With respect to the N data (L, K), binarization is performed using the template "value: 1 number of data". This binarization is “binary conversion”. Binarization is performed as follows. That is, N pieces of data (L, K) are arranged in the order of the density, and the density data of pixels from the density level: 256 side to Σn (j) th is “1”,
The density data of the remaining pixels is set to "0".

The binary converted N pixel data are arranged in the original address order to form a comparison pattern. As will be readily understood from the above description, the number of values: 1 in the binary template and the comparison pattern created by cutting out from the read circular pattern are:
If the number of value: 1 and the number of value: 0 are equal to each other, and the circular pattern that is the source of the template and the circular pattern that is the identification target are the same pattern,
The agreement is extremely high.

Further, since the comparison pattern is standardized under the same conditions as the template by the above binary conversion, even if there is an individual difference in the surface state of coins or game yen bills, the similarity or dissimilarity calculation is performed. Can be reliably performed.

[0027]

EXAMPLES Specific examples will be described below.

FIG. 2 is a diagram for explaining an embodiment of the circular pattern identifying method according to the present invention. As shown in FIG. 2 (I), the circular pattern to be identified is a pattern of a 500-yen coin. FIG. 2 (II) shows a state in which the pattern of (I) is read and the density of each pixel is stored as a multivalued digital signal. In this state, the center coordinates of the circular pattern are detected (claim 2).

Various well-known methods can be used for the central coordinate detection algorithm. Here, the central coordinate (x ₀ , x ₀ ,
y ₀ ). As described above, when reading a circular pattern, if the position of the 500-yen coin is fixed at a fixed position, the center coordinates (x ₀ , y ₀ ) can be given in advance, and the center coordinates cannot be detected. Useless.

Then, as shown in FIG. 2 (III), a "ring portion surrounded by concentric circles" at a predetermined distance from the center coordinates of the circular pattern is cut out, and this is cut out as shown in FIG. 2 (I).
As shown in V), it is developed into a "two-dimensional array" in the X and Y directions. The state thus developed is the "two-dimensional basic data".

In this example, the basic data is composed of 512 and 22 data in the X and Y directions, respectively. Y
The direction corresponds to the “radial direction” of the circle diameter pattern. Also,
Both ends in the X direction are the reference line Z in FIG. 2 (III) (the direction is appropriately set in the image memory unit, and when the center coordinate is determined, it is uniquely determined by the position of the center coordinate and the direction. ) Corresponds to.

As described above, the cut-out two-dimensional basic data can be used as it is as the "identification data array" (claim 3), but in this embodiment, the "differential absolute value" is further added. A "sum creation" process and a "data compression" process are performed (claim 4).

First, regarding each element of the "two-dimensional basic data" shown in FIG. 1 (IV), the difference from the data of the "eight surrounding elements" of the element is calculated in each of the X and Y directions. Ask.
FIG. 2 (V) shows the “difference in the X direction” of one element (element filled in black) of the basic data shown in FIG. 2 (IV) with respect to eight surrounding elements. Further, FIG. 1 (VI) shows the “difference in the Y direction” for the same eight elements and the surrounding eight elements. These differences are called “differential values”. The sum of the absolute values of the differential value in the X direction and the differential value in the Y direction is calculated, and the calculation result is stored as the “sum of differential absolute values” for the element.

FIG. 2 (VII) shows a data array of differential absolute value sums stored in this way. Figure 1
In the (IV) X direction, the η direction also corresponds to the Y direction. In the basic data, the element row at both ends in the Y direction has only five surrounding elements, and the differential value cannot be calculated.
In the differential absolute value sum data of (VII), the number of elements in the η direction is 20. When the differential absolute value sum generation process is performed, the density change amount in the basic data can be converted into data, so that the change in the unevenness of the pattern image can be emphasized.

Next, the data array shown in FIG. 2 (VII) is divided into data groups (one of which is shown by hatching) as shown in FIG. 2 (VIII), and data (differentiation) in each data group is divided. The sum of absolute values) is taken and this is newly set as "aggregated data" to form a data array (FIG. 2 (IX)). The data array of the aggregated data thus obtained is the "identification data array". In this embodiment, the “data group” has 4 elements in the ξ direction and 2 elements in the η direction.
Therefore, the identification data array is composed of 128 elements corresponding to the ξ direction: 128 elements in the u direction, and 10 elements in the v direction corresponding to the η direction. By performing the "data compression" process in this way, the time required for the matching calculation can be effectively shortened.

The binary template has the same size as the above-mentioned identification data array, and the same procedure as that for forming the above-mentioned "identification data array" from the standard pattern of 500-yen coins, which is the identification pattern. By binarizing the aggregated data obtained through the above-mentioned “P-tile method”,
It is preformed.

In FIG. 3, (a) shows an example in which a 10 * 128 data aggregated as a template is made into a histogram according to the level of the aggregated data.
Binarization is a histogram (the area of the histogram is 10 ×
128), the shaded area is the total area (1
By setting a value of 1 for data having a level higher than a threshold level (represented by a chain line) selected so as to have a predetermined ratio (for example, 30% = 384) of 0 × 128, and setting a value of 0 for other data. To do.

FIG. 3B is an explanatory diagram showing the state of a binary template obtained by arranging the binarized data in this manner according to the aggregated data array. Value: 1 for black areas, value: 0 for other areas
Is. The number of data with a value of 1 is 38 in the above example.
There are 4 data.

With respect to the "identification data array (FIG. 2 (IX))" formed as described above, the above "binary conversion" is performed.
To form a “comparison pattern”. That is, 384 data are selected in the descending order of the data level in the identification data array, and these are set to a value of 1 and the other data are set to a value of 0. By arranging the binarized data in this manner according to the original identification data array, the data shown in FIG.
A “comparison pattern” obtained by binarizing the identification data array shown in X) is obtained. The number of data of value: 1 in the comparison pattern is equal to the number of data of value: 1 in the template.

The comparison pattern can be expressed as a binary function: F (u, v) with u and v as variables, and one template also has a binary function: T (u, v with u and v as variables.
v) can be represented. The binary function takes only 0 and 1 as the function value.

Next, a calculation step is performed to calculate the similarity or dissimilarity between the comparison pattern and the template. As shown in FIG. 4, for example, in the case of calculating the “dissimilarity”, the calculation formulas: ΣΣ | ft− and ΣΣ (f−
t) ² , when calculating the “similarity”, ΣΣ (f ·
t) (the sum is taken over the entire range of addresses u and v) can be used as an arithmetic expression.

The matching operation for calculating the similarity or dissimilarity is 0 to 2 for the angular coordinates of the circular pattern.
Perform up to π. This is because a matching error due to rotational symmetry of the circular pattern with respect to the center coordinates is removed. To do this, in the comparison pattern u = i (i = 0 ...
The process of converting the element of 126) to u = i + 1, converting the element of u = 127 to u = 0, and generating the “new comparison pattern” cyclically may be repeated 128 times in total. The true "dissimilarity" is given by the "minimum value" of the result of this series of operations. The true similarity is given by the "maximum value" of the result of this series of operations.

The dissimilarity (similarity) thus obtained is compared with a threshold value of the dissimilarity (similarity) statistically obtained in advance for the identification pattern, and the dissimilarity (similarity) is determined. If it is higher (lower) than the threshold value, it is determined that the read circular pattern does not include a pattern corresponding to the template (identification step).

FIG. 5 is a functional diagram showing an example of an apparatus for carrying out the present invention. The circular pattern is read as a video signal by the image input unit 1 and converted into a digital signal by the A / D conversion unit 2. The image input unit 1 and the A / D conversion unit 2 form an “input unit”.

The data input from the input section is stored in the image memory section 3. Then, the "center coordinates" of the stored circular pattern are detected by the center coordinate calculation unit 4, and the "ring-shaped area" is centered on the detected center coordinates.
Is cut out by the cutout unit 5. The image cutout unit 5 also extracts “two-dimensional basic data” from the cutout information.
Is generated and stored in the built-in memory of the clipping unit 5. Based on the stored contents of the built-in memory, the differential absolute value sum creation unit 6 forms a “differential absolute value sum data array”,
The result is output to the pixel aggregation processing unit 7.

The pixel aggregation processing unit 7 forms an “identification data array” from the differential absolute value sum data array and stores it in the memory incorporated in the image aggregation processing unit 7. Based on the stored contents of this memory, the concentration order table creating unit 8
A table required for value conversion (a table in which data is arranged in order of level) is formed.

In the binary conversion unit 9, binary conversion is performed based on the table formed by the density order table creation unit 8 and the data of the template value: 1 stored in the memory unit 12. The comparison pattern is generated. The differential absolute value sum creation unit 6, the pixel aggregation unit 7, the density order table creation unit 8, and the binary conversion unit 9 form a “data processing unit”.

The dissimilarity calculation unit 10 serves as a "calculation unit".
The matching calculation between the comparison pattern and the template (stored in the memory unit 13) is performed from 0 to 2π with respect to the angular coordinates of the circular pattern, and the “dissimilarity” between the comparison pattern and the template is detected.

The control identification section 11 determines "whether or not the circular pattern includes a pattern corresponding to the template" based on the above detection result. The control identification unit 11 performs identification determination and controls each unit so that each process from the detection of the central coordinates to the calculation of the dissimilarity is properly performed.

The part excluding the input part of the above device can be configured by a computer or a computer and an additional memory.

FIG. 6 is a functional diagram showing an example of an apparatus for carrying out the invention described in claim 5. In order to avoid complication, the same symbols as those in FIG.

In this device, the memory 15 of the memory section
, A plurality of shape data (data of a plurality of concentric ring-shaped areas) as a cutout shape to be cut out, a plurality of binary template data corresponding to each cutout shape, and a value in these templates: 1
Are stored in the memories 13A and 12A, respectively.

The process from the cutout of the image to the calculation of the dissimilarity is repeated sequentially while changing the cutout region.
The cutout area is changed by the cutout data switching unit 14
The image cut-out unit 5 is controlled by means of, and the template is switched every time the area to be cut out is changed, and the template switching unit 17 is used to switch the data of the value of 1 and the data switching unit of the value of 1. Done by.

As described above, by performing matching with two or more cutout portions, the accuracy of identification is significantly improved.

[0055]

As described above, according to the present invention, a novel "circular pattern identification method" can be provided. According to the present invention, only the "characteristic portion" is cut out rather than the entire circular pattern, and the identification is performed in relation to the binary conversion. Therefore, it is possible to effectively reduce the operation time of the matching operation necessary for the determination of the unlikeness. If data compression is performed as in the method of claim 4, the calculation time can be further shortened.

The value of the template: 1
Since it is performed between the comparison pattern and the binary template that are standardized by the number of data, the influence of the contrast difference between the individual images of the circular patterns is small, and the calculation is simple and the calculation time can be shortened.

According to the fifth aspect of the present invention, the number of parts to be collated increases, so that the identification rate can be significantly improved.

[Brief description of drawings]

FIG. 1 is a diagram for explaining the present invention.

FIG. 2 is a diagram for explaining from a reading process to an identification data array in the embodiment of the method according to claim 4;

FIG. 3 is a diagram for explaining a binary template.

FIG. 4 is a diagram for explaining an analogy calculation.

FIG. 5: One of the devices for carrying out the method according to claim 4
It is a figure which shows an example as a functional diagram.

FIG. 6 a device for carrying out the method according to claim 5;
It is a figure which shows an example as a functional diagram.

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Office reference number FI technical display location 9061-5L 460 A 5

Claims (5)

[Claims] 1. An input step of reading a circular pattern and storing the density of each pixel as a digital signal, and a portion which shows the characteristics of the circular pattern well from the stored information, with the center coordinates of the circular pattern as the center. A cutting step of obtaining a two-dimensional basic data by cutting out into a circular shape or a ring shape and performing necessary processing on this basic data to obtain a data array for identification having the same size as the binary template; The data array is binary-converted by using the number of data of the template: 1 to obtain a comparison pattern, and a binary conversion step of obtaining a comparison pattern, and a matching calculation between the comparison pattern and the template are performed on the angular coordinates of the circular pattern. Is performed from 0 to 2π, and a calculation step of detecting the similarity or dissimilarity between the comparison pattern and the template is detected. Based on the similarity score or dissimilarity, the circular pattern, and having a determining identification step whether including a pattern corresponding to the template, a circular pattern identification method. 2. The circular pattern identifying method according to claim 1, wherein the center coordinates of the circular pattern of the information stored in the input step are detected in the cutting step. 3. The circular pattern identification method according to claim 1, wherein the binary template has the same size as the two-dimensional basic data, and the basic data itself is used as an identification data array. The circular pattern identification method. 4. The circular pattern identification method according to claim 1 or 2, wherein the processing for obtaining the identification data array from the two-dimensional basic data includes differential absolute value sum creation processing and / or data compression processing. A method for identifying a circular pattern, characterized by: 5. The circular pattern identification method according to claim 1, 2 or 3 or 4, wherein a plurality of binary templates are prepared, and a plurality of types of identification data arrays are associated with each template in a one-to-one correspondence. A circular pattern identification method characterized by detecting the similarity or dissimilarity of a template corresponding to a comparison pattern obtained for each identification data array.