JP3631618B2 - Discrimination method of circular pattern - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a circular pattern discrimination method. More specifically, the present invention relates to a method for discriminating a circular pattern suitable for discriminating coins and the like.
[0002]
[Prior art]
In a circular pattern recognition device used for coin discrimination or the like, as shown in Japanese Patent Laid-Open No. 5-143731, when a recognition object is taken into the device, a circular pattern of the recognition object is photographed and used as a basis. A two-dimensional comparison pattern is created. In order to create a comparison pattern from a circular pattern, for example, a circular pattern or an annular shape is cut out around the center coordinates of the circular pattern, and necessary processing is performed on the cut out digital data.
[0003]
Then, the created comparison pattern is compared with a template of various coins prepared in advance, that is, a reference pattern, and a similarity or dissimilarity between the two patterns is calculated (hereinafter referred to as a matching calculation). This matching operation is performed while rotating the relative angle between the comparison pattern and the template from 0 to 2π, for example, in 128 steps with respect to the angular coordinates of the circular pattern. An example of the result of calculating the similarity is shown in FIG.
[0004]
Furthermore, the circular pattern is recognized and discriminated based on the matching value that is the result of the matching operation. The determination of similarity is made using the value of the highest point P1 (hereinafter referred to as the best matching value) among the matching values obtained at each stage from 0 to 2π, and the best matching value exceeds a predetermined threshold value M0. It is judged by whether or not. That is, if the best matching value related to the similarity exceeds the threshold M0, it is determined that the comparison pattern and the template are likely to match, and if the best matching value related to the dissimilarity exceeds the threshold, the comparison is made. It is determined that there is a high possibility that the pattern and the template do not match.
[0005]
[Problems to be solved by the invention]
However, in the above-described circular pattern discrimination method, determination is performed using only one best matching value, so even if the comparison pattern and the template are completely different types, the best matching value of the comparison pattern is accidentally changed. If the threshold value is also exceeded, there is a risk that the comparison pattern and the template are likely to match. If the comparison pattern and the template are different types, for example, as shown in FIG. 5, the waveform (solid line) of the relationship between the rotation angle and the similarity when the template is matched with the template and the rotation of the comparison pattern with respect to the template Although the waveform is completely different from the waveform of the relationship between the angle and the similarity (the one-dot chain line), if the best matching value of the comparison pattern exceeds the threshold value M0, it is determined that the comparison pattern and the template are highly likely to match. End up.
[0006]
If there is such an erroneous determination, the recognition accuracy becomes low, and even if it is a false coin that should originally be excluded, it may be recognized as an appropriate coin.
[0007]
Accordingly, an object of the present invention is to provide a circular pattern discrimination method capable of improving recognition accuracy.
[0008]
[Means for Solving the Problems]
In order to achieve this object, the invention described in claim 1 obtains a two-dimensional comparison pattern from pattern data cut out in a circular shape from a circular pattern, and performs a matching operation with a template prepared in advance on the comparison pattern. In the circular pattern discrimination method in which the degree of similarity is obtained by rotating the angular coordinates from 0 to 2π and the circular pattern is discriminated, a threshold value for the degree of similarity for discriminating the circular pattern is determined in advance. , When at least one of the similarities when rotated from 0 to 2π exceeds the threshold, the similarity at other angular coordinate positions predetermined corresponding to the template is compared with the threshold for the similarity, The circular pattern is discriminated based on the result.
[0009]
Therefore, in addition to at least one similarity exceeding the threshold, the similarity at other angular coordinate positions is also compared with the threshold, so that the accuracy of discriminating the circular pattern can be determined by comparing a plurality of similarities with the threshold. Can be improved. That is, according to this determination method, even if the comparison pattern and the template are completely different types and one similarity accidentally exceeds the threshold, the comparison pattern and the template are similar to each other as in the past. It is possible to appropriately determine that the dissimilarity is not based on the fact that the similarity at other angular coordinate positions does not reach the threshold value.
[0010]
Further, in the circular pattern discrimination method according to claim 2, the other angular coordinate positions determined in advance corresponding to the template are set in the vicinity where the similarity is the peak position or a peak position different from the peak position. It is made to become.
[0011]
Therefore, since a plurality of portions having similarities can be taken out and used for determination, the determination accuracy can be improved.
[0012]
Furthermore, the invention described in claim 3 obtains a two-dimensional comparison pattern from pattern data cut out in a circular shape from a circular pattern, and performs a comparison operation with a template prepared in advance for the comparison pattern from 0 with respect to the angular coordinates of the circular pattern. In the circular pattern discriminating method in which the degree of similarity is obtained by rotating to 2π to discriminate the circular pattern, n similarities obtained by relatively rotating the comparison pattern and the template from 0 to 2π. The n-dimensional vector or waveform is obtained by arranging the degrees according to the rotation angle, and this is used as the n-dimensional target vector or target waveform. Separately, the template and the template (templates) are rotated relative to the angular coordinates from 0 to 2π. The n similarities obtained in this way are arranged according to the rotation angle to obtain an n-dimensional vector or waveform. This is used as an n-dimensional reference vector or reference waveform, and the circular pattern is discriminated by obtaining the correlation between the n-dimensional target vector and the n-dimensional reference vector, or the correlation between the target waveform and the reference waveform. I have to.
[0013]
Therefore, since the n-dimensional vector or waveform is a continuous transition of the similarity, the circular pattern is discriminated based on the correlation between the n-dimensional vector or the waveform of the target and the reference as in the conventional case. The accuracy of discriminating a circular pattern can be improved as compared with the case where only one similarity is compared with a threshold value. That is, instead of the conventional local evaluation of the similarity, for example, the shape of a waveform and the tendency of transition can be used as the discrimination information, so that the discrimination accuracy can be improved.
[0014]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, the configuration of the present invention will be described in detail based on an example of an embodiment shown in the drawings. FIG. 2 shows an embodiment of a circular pattern discriminating apparatus 1 that uses the circular pattern discriminating method of the present invention. In the present embodiment, the circular pattern discriminating device 1 is used as a coin 2 discriminating device. The circular pattern discriminating apparatus 1 has an image input unit 3 for photographing a coin 2 and an identification processing unit 4 for discriminating a photographed circular pattern.
[0015]
The image input unit 3 includes a light emitting element 5 that irradiates the surface of the coin 2, an optical system 6 that receives reflected light from the coin 2 and reads an image of the coin 2 as a circular pattern, and a circular pattern that passes through the optical system 6. It has an image sensor 7 such as a CCD that converts it into an electrical signal. The identification processing unit 4 processes an A / D conversion element 8 for digitally converting an analog signal of a circular pattern from the image sensor 7, a storage element 9 for storing the digitally converted circular pattern, and a stored circular pattern. And a determination unit 10 for determining the coin 2.
[0016]
As an operation procedure of the circular pattern discriminating apparatus 1, first, an image input process is performed by picking up an image of the coin 2 by the image input unit 3. The circular pattern is digitally converted by the A / D conversion element 8 and stored in the storage element 9.
[0017]
Further, in the discriminating unit 10, for example, a circular pattern or an annular shape is cut out around the center coordinates of the circular pattern, and necessary processing is performed on the cut out digital data to create a comparison pattern. Then, the created comparison pattern is compared with templates of various coins 2 prepared in advance.
[0018]
Here, a procedure for comparing the comparison pattern and the template to determine the coin 2 will be described based on the flowchart shown in FIG. 1 and FIG.
[0019]
First, a matching operation is performed on the similarity between the comparison pattern and the template (step 1). This matching operation is performed while rotating the relative angle between the comparison pattern and the template from 0 to 2π, for example, in 128 steps with respect to the angular coordinates of the circular pattern. An example of the result of this matching operation is indicated by ● in FIG. In this embodiment, the relative angle is rotated once in 128 steps, but it is needless to say that the relative angle is not limited to 128 steps.
[0020]
Then, it is determined whether the matching value of the point Pb having the highest similarity, that is, the best matching value exceeds a predetermined threshold value Mb (step 2). That is, it is determined whether at least one of the matching values exceeds the threshold value Mb. Here, those satisfying the conditions are selected and become denomination candidates.
[0021]
On the other hand, for each template, a waveform of the relationship between the rotation angle and the similarity when the template is matched with the template is obtained in advance (shown by a solid line in FIG. 3). Then, a threshold value at a predetermined rotation angle is set in consideration of a characteristic shape such as a peak of the waveform.
[0022]
Further, for the selected candidate, the matching value at another angular coordinate position determined in advance corresponding to the template is compared with a threshold value (step 3). The threshold value here may be the same as or different from the threshold value used in step 2, or may be different according to the characteristics of each denomination. In the example shown in FIG. 3, the rotation angle θ1 that becomes the other peak points P1 and P2 with reference to the rotation angle θb that becomes the best matching value so that a plurality of peak points P1 and P2 that are characteristic points of the curve can be detected. And threshold values M1 and M2 are set for the matching values at θ2 and θ2. Then, it is determined whether the matching values at the rotation angles θ1 and θ2 exceed the threshold values M1 and M2.
[0023]
If the matching value exceeds the threshold value in both step 2 and step 3, it is determined that the circular pattern is that of the appropriate coin 2, and the coin 2 is accepted (step 4). On the other hand, if the matching value does not exceed the threshold value in step 2 or step 3, it is determined that the circular pattern is of an improper coin 2 and the coin 2 is excluded (step 5). This procedure is performed for each candidate template.
[0024]
As described above, according to the circular pattern discriminating apparatus of the present embodiment, in addition to at least one matching value exceeding the threshold value, the matching values at other angular coordinate positions are also compared with the threshold value. By comparing the matching value with a threshold value, the accuracy of discriminating the circular pattern can be improved. That is, since there are a plurality of evaluation points that are information amounts for discrimination, it is possible to improve the reliability of the discrimination results as compared with the case where the evaluation is performed with only one point that is the best matching value as in the prior art.
[0025]
Further, since the best matching value is compared with the threshold value Mb and the other matching values are compared with the threshold values M1 and M2 so that the peak points P1 and P2 which are characteristic points of the curve can be detected, the accuracy of the determination is further increased. Can be improved.
[0026]
The above-described embodiment is an example of a preferred embodiment of the present invention, but is not limited thereto, and various modifications can be made without departing from the scope of the present invention. For example, in the present embodiment, threshold values M1 and M2 other than the threshold value Mb for the best matching value are set so as to detect characteristic peak points P1 and P2, but the present invention is not limited to this. For example, a peak including the best matching value is included. It may be determined whether the threshold value Mb is exceeded for several steps (or several degrees) before and after the best matching value. Also in this case, since a plurality of characteristic portions can be taken out and used for determination, the determination accuracy can be improved.
[0027]
In the above-described embodiment, as a procedure for comparing the comparison pattern and the template to determine the coin 2, in addition to at least one matching value exceeding the threshold value, a matching value at another angular coordinate position is also compared with the threshold value. However, it is not limited to this.
[0028]
For example, as shown in the flowchart of FIG. 4, first, a matching calculation is performed on the similarity between the comparison pattern and the template (step 6). This matching operation is performed while rotating the relative angle between the comparison pattern and the template from 0 to 2π with respect to the angular coordinates of the circular pattern, for example, in n = 128 steps. Here, n matching values obtained by rotating from 0 to 2π are arranged according to the rotation angle to obtain the target n-dimensional vector or waveform S. In the present embodiment, the relative angle is rotated once at n = 128, but n is not limited to 128.
[0029]
Then, a reference n-dimensional vector or waveform St prepared in advance by matching the template with the template is compared with the target n-dimensional vector or waveform S to obtain a correlation (step 7). The correlation calculation method here is not particularly limited, and a known or new method for evaluating the similarity or dissimilarity of two shapes can be used.
[0030]
As a result, if the correlation value between the two n-dimensional vectors of the target and the reference or the waveforms S and St is larger than the threshold value, it is determined as an appropriate coin 2 (step 9), and if it is smaller than the threshold value, it is determined as an inappropriate coin 2. (Step 10).
[0031]
According to this method, since the n-dimensional vector or waveform is a continuous transition of the similarity, the accuracy of discriminating the circular pattern is improved as compared with the conventional case where only one similarity is compared with the threshold value. Can be made. That is, instead of the conventional local evaluation of the similarity, for example, the shape of a waveform and the tendency of transition can be used as the discrimination information, so that the discrimination accuracy can be improved.
[0032]
In each embodiment described above, the similarity is calculated as the matching calculation. However, the present invention is not limited to this, and the similarity may be calculated based on the similarity.
[0033]
Furthermore, in each of the embodiments described above, the circular pattern discrimination method is used in the coin 2 discriminating device, but the present invention is not limited to this and can be widely applied to circular pattern recognition. For example, it can be used to recognize various coins, crowns, labels, fingerprints, irises, and the like. Even in that case, highly accurate image recognition can be performed.
[0034]
【The invention's effect】
As is clear from the above description, according to the method for discriminating a circular pattern according to claim 1, in addition to at least one similarity exceeding the threshold, the similarity at other angular coordinate positions is compared with the threshold. Therefore, the accuracy of discriminating a circular pattern can be improved by comparing a plurality of similarities with a threshold value. That is, since there are a plurality of evaluation points that are information amounts for discrimination, it is possible to improve the reliability of the discrimination results as compared with the case where the evaluation is performed with only one point that is the best matching value as in the prior art.
[0035]
In addition, according to the method for discriminating a circular pattern according to the second aspect, it is possible to take out a plurality of characteristic parts having similarities and use them for the determination.
[0036]
Furthermore, according to the method for discriminating a circular pattern according to claim 3, since the n-dimensional vectors or waveforms that are continuous transitions of the similarity are compared, in the conventional local evaluation of the similarity, For example, the shape of a waveform, the tendency of transition, etc. can be used as discrimination information. For this reason, the accuracy of discrimination can be improved.
[Brief description of the drawings]
FIG. 1 is a flowchart showing one embodiment of a circular pattern discrimination method of the present invention.
FIG. 2 is a block diagram showing an embodiment of a discriminating apparatus that uses a discriminating method of a circular pattern.
FIG. 3 is a graph showing an example of a relationship between a rotation angle and a similarity when a comparison pattern and a template are subjected to a matching operation on the template.
FIG. 4 is a flowchart showing another embodiment of a circular pattern discrimination method.
FIG. 5 is a graph showing an example of a relationship between a rotation angle and a similarity when a comparison pattern and a template are subjected to a matching operation with respect to a template by a conventional discrimination method.
[Explanation of symbols]
1 Circular pattern discrimination device

Claims (3)

A two-dimensional comparison pattern is obtained from pattern data cut into a circular shape from a circular pattern, and the comparison pattern is prepared in advance and a matching operation is performed by rotating from 0 to 2π with respect to the angular coordinates of the circular pattern. In the circular pattern determination method for determining the circular pattern, a threshold value for the similarity for determining the circular pattern is determined in advance, and the rotation is performed from 0 to 2π. When at least one of the similarities exceeds the threshold, the similarity at other angular coordinate positions predetermined corresponding to the template is compared with the threshold for similarity, and based on both results, the circular pattern A method of discriminating a circular pattern, characterized in that discrimination is performed. The other angular coordinate position predetermined corresponding to the template is set in the vicinity where the similarity is a peak position or a peak position different from the peak position. Discrimination method of circular pattern. A two-dimensional comparison pattern is obtained from pattern data cut into a circular shape from a circular pattern, and the comparison pattern is prepared in advance and a matching operation is performed by rotating from 0 to 2π with respect to the angular coordinates of the circular pattern. In the circular pattern discriminating method for discriminating the circular pattern, n similarities obtained by relatively rotating the comparison pattern and the template from 0 to 2π are arranged according to the rotation angle. An n-dimensional vector or waveform is obtained and used as an n-dimensional object vector or object waveform, and n similarities obtained by rotating the template and the template relatively from 0 to 2π with respect to angular coordinates are obtained. Arranging according to the rotation angle, obtaining an n-dimensional vector or waveform and converting it to an n-dimensional reference vector Is a reference waveform, and a circular pattern is discriminated by determining the correlation between the n-dimensional target vector and the n-dimensional reference vector, or the correlation between the target waveform and the reference waveform. Method for discriminating circular patterns characterized by

Images