JP2000048197A - Method for identifying circular pattern, and device therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make a circular pattern identifying device possible to be compact and inexpensive, and to shorten the processing time required for the pre- processing process of a matching process. SOLUTION: A coordinate transformation is applied to a circular pattern so that a comparison pattern can be formed, and the comparison pattern is compared with a template being a reference so that the circular pattern can be identified. In this method, a feature emphasizing process 22a for emphasizing the features of the circular pattern is applied, and then a data compression is applied (a data compression process 22b), and a coordinate transformation is applied to the circular pattern after the data compression (polar coordinate transformation process 22c), and a comparison pattern is formed (a concentration conversion process 23), and the comparison pattern is compared with a template being a reference so that the circular pattern can be identified (a matching process 24).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a method and an apparatus for identifying a circular pattern. More specifically, the present invention relates to a pre-process for forming data of the same size as a template from a circular pattern in order to identify a circular pattern, that is, a pre-process of a matching process.

[0002]

2. Description of the Related Art As a pattern identification method for a circular object such as a coin, for example, a characteristic portion is extracted from an image of an object, a comparison pattern is generated from the identification pattern, and reference data created in advance is created. A template matching method for comparing with (template) is known.
As shown in FIG. 6, the pattern identification method of the template matching method includes an image input step 101, a pre-processing step 102, a density conversion step 103, and a matching step 104.
The processing is performed in the order of
The pre-processing step 102 for generating the identification pattern has been performed in the order of the polar coordinate conversion step 102a, the feature enhancement step 102b, and the data compression step 102c.

That is, first, a polar coordinate conversion step 102a
In FIG. 7, the characteristic portion is cut out from the image of the object to be identified into a ring shape (in FIG. 7A, a range surrounded by two dashed lines 105 and 106), subjected to polar coordinate transformation, and transformed into a rectangular matrix format (identification). (Pattern) (FIG. 7B).
Note that the hatched portions in the figure illustrate pattern patterns on the surface of the identification target. In this polar coordinate conversion, the coordinates of a point (pixel) to be extracted from the image data are stored in advance in a memory (coordinate conversion table) as position information,
At the time of execution, the position data is read from the memory, the image data on the image corresponding to the data is taken out, and a matrix type array is generated.

Thereafter, in a feature emphasizing step 102b, differentiation processing of the coordinate conversion table is performed to emphasize the edge of the identification pattern (FIG. 7 (C)).
02c is performed to compress the identification pattern with a plurality of pixels as one unit (FIG. 7D).

[0005]

However, in the conventional pattern identification method, the preprocessing step 102 is performed by a polar coordinate conversion step 102a → a feature enhancement step 102b → a data compression step 1.
Since it is performed in the order of 02c, the polar coordinate conversion step 102a is performed on the original image data before compression. For this reason, the amount of position information increases in accordance with the number of pixels to be cut out from the image data, and a large-capacity memory (coordinate conversion table) is required to store the position information. And cost reduction are difficult,
Since a large amount of data is handled for polar coordinate conversion, it has been difficult to reduce the processing time.

SUMMARY OF THE INVENTION An object of the present invention is to provide a method and an apparatus for identifying a circular pattern, which can reduce the size, cost and processing time of the apparatus.

[0007]

According to the first aspect of the present invention, a circular pattern is subjected to coordinate transformation to form a comparison pattern, and the comparison pattern is compared with a reference template. In the method of identifying a circular pattern in which a circular pattern is identified, data compression is performed after performing a feature emphasizing step of enhancing characteristics of the circular pattern, and coordinate conversion is performed on the circular pattern after the data compression, and the comparison is performed. A circular pattern is identified by forming a pattern and comparing the comparison pattern with a reference template.

Therefore, the pre-processing step of the method for identifying a circular pattern is constituted in the order of the feature emphasis step, the data compression step, and the polar coordinate conversion step. That is, the data after compression is subjected to coordinate conversion, and the amount of data handled in the polar coordinate conversion process is reduced, thereby reducing the required memory capacity.

In this case, as in the method for identifying a circular pattern according to the second aspect, the circular pattern is obtained by reading an image to be identified and forming a circular ring shape from image information in which the density of each pixel is stored as a digital signal. It is desirable that the feature emphasizing process is a process of emphasizing a feature by differential processing. By doing so, a pre-processing step of a method for identifying a circular pattern suitable for identifying a coin or the like is provided.

According to a third aspect of the present invention, there is provided an apparatus for identifying a circular pattern, comprising: an image input unit for reading an image to be identified; an image memory for storing the density of each pixel of the image to be identified read by the image input unit; A feature emphasis unit for emphasizing the density feature of the identification target image, a data compression unit for compressing the identification target image emphasized by the feature emphasis unit, and pattern identification from the identification target image compressed by the data compression unit. The image processing apparatus includes an image cutout unit that cuts out a circular region to be performed, and a matching calculation unit that compares the region cutout by the image cutout unit with a previously created reference template.

Therefore, after the image to be identified read by the image input unit is stored as a digital signal in the image memory, the feature of the image to be identified is enhanced by the feature emphasizing unit. Thereafter, the identification target image is compressed by the data compression unit, and a predetermined area is cut out by the image cutout unit to form an identification pattern. The similarity or dissimilarity is calculated in the matching calculation unit based on the identification pattern. That is, the data after compression is subjected to coordinate transformation, and the amount of data handled when forming an identification pattern in the image cutout unit is reduced, so that the required memory capacity is reduced.

In this case, as in the circular pattern discriminating apparatus according to the fourth aspect, the image cutout section has a coordinate conversion reference table, and performs coordinate conversion of the circular area to convert the discrimination pattern of the same size as the template. It is desirable to cut it out. Here, the template is a rectangular data array. Therefore, the image cutout unit performs coordinate conversion of the image data of the circular area into a rectangular data array using the position information of the coordinate conversion reference table to form an identification pattern.
Also in this case, the data after compression is subjected to coordinate conversion, and a small-sized coordinate conversion lookup table is sufficient.

[0013]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The configuration of the present invention will be described below in detail based on the best mode shown in the drawings.

FIGS. 1 and 3 show an example of an embodiment of a circular pattern identification method to which the present invention is applied, and FIG. 2 shows an example of an embodiment of a circular pattern identification apparatus to which the present invention is applied. 2 includes an image input unit 1 for reading an image 37 to be identified, an image memory 3 for storing the density of each pixel of the image 37 for identification read by the image input unit 1, an image for identification 3 A feature emphasizing unit 4 for emphasizing the density feature of 37; and a data compressing unit 5 for compressing the identification target image 37 emphasized by the feature emphasizing unit 4.
And the identification target image 37 compressed by the data compression unit 5
Area (circular pattern) 3 for pattern identification from
The image processing apparatus includes an image extraction unit 6 that extracts the image 6 and a matching operation unit 8 that compares a region extracted by the image extraction unit 6 with a template 12 that is a reference created in advance. Further, the image cutout unit 6 has a coordinate conversion reference table 10, and performs coordinate conversion on the cutout circular area 36 to cut out an identification pattern 35 having the same size as the template 12.

In the present embodiment, the image signal of the identification target imaged by the image input unit 1 is converted into digital image data by the A / D conversion unit 2 and stored in the image memory 3. Further, a density conversion unit 7 is provided between the image cutout unit 6 and the matching calculation unit 8, and the density conversion unit 7 uses the density conversion reference table 11 to identify the identification pattern 3.
5, a comparison pattern 38 is formed (FIG. 5). Then, the matching calculation unit 8 compares the comparison pattern 38 with the template 12, and calculates the similarity or the dissimilarity of the comparison pattern 38. Further, the control / identification unit 9
In, the identification target is determined based on the result obtained by the matching calculation unit 8.

Next, a method for identifying a circular pattern to which the present invention is applied will be described with reference to FIG. The method of identifying a circular pattern is to perform a coordinate conversion on a circular pattern (circular region) 36 to form a comparison pattern 38 and compare the comparison pattern 38 with the reference template 12 to identify the circular pattern 36. After performing a feature emphasizing step 22a for emphasizing the features of the circular pattern 36, data compression is performed (data compression step 22b), and coordinate conversion is performed on the circular pattern 36 after data compression (polar coordinate conversion step 22c). A comparison pattern 38 is formed (density conversion step 23), and the circular pattern 36 is identified by comparing the comparison pattern 38 with the reference template 12 (matching step 24). In the present embodiment, the discriminating circular pattern 36 is a circular ring-shaped pattern in image information in which the discrimination target image 37 is read and the density of each pixel is stored as a digital signal.

More specifically, in this method for identifying a circular pattern, processing is performed in the order of image input step 21, preprocessing step 22, density conversion step 23, and matching step 24. In preprocessing step 22, feature enhancement is performed. Processing is performed in the order of step 22a → data compression step 22b → polar coordinate conversion step 22c. That is, in the circular pattern identification method of the present invention, unlike the conventional pattern identification method shown in FIG. 6, polar coordinate transformation is performed on an image after data compression. The image input process 21, the density conversion process 23, and the matching process 24
Performs the same processing as the conventional pattern identification method.
A detailed description of these will be omitted.

First, in the first image input step 21, the image to be identified 37 is read by the image input unit 1, and the density of each pixel is converted into a digital signal by the A / D converter 2. FIG. 3A conceptually shows the read identification target image 37. In the figure, a solid circle 31 exemplifies the contour of the identification target such as a coin, and a hatched square portion 32 exemplarily shows a pattern pattern on the surface of the identification target. Also, two 2
Dashed-dotted circles 33 and 34 indicate the range of image data to be extracted, and indicate that a ring-shaped portion sandwiched between two two-dot-dashed circles 33 and 34 is cut out. That is, the range sandwiched between the circles 33 and 34 is the circular pattern 36.
The centers of the circles 33 and 34 of each two-dot chain line coincide with the center of the identification target, and the same range can always be cut out even if the direction (rotation angle) of the identification target is different.

Next, the feature emphasizing step 22 of the preprocessing step 22
In a, the feature emphasizing section 4 performs a process of emphasizing the features of the pattern (FIG. 3B). There are various methods for this feature emphasis processing. For example, edge emphasis processing for emphasizing features by differential processing for extracting a change amount is used. After that, the data compression section 22 performs a data compression step 22b, and compresses the image data with the emphasized features in units of several pixels (FIG. 3C).

Next, a polar coordinate conversion step 22c is performed in the image extracting section 6, and data of a ring-shaped portion (circular pattern 36) sandwiched between two two-dot chain lines 33 and 34 is cut out to obtain rectangular data. The array is converted into an identification pattern 35 having the same format as that of the template 12 stored in advance (FIG. 3D). Specifically, as shown in FIG. 4, coordinate data of a pixel to be cut out is obtained in advance by calculation, and this is tabulated to obtain a coordinate conversion reference table 10.
Is created, and coordinate conversion is performed with reference to the coordinate conversion reference table 10. This coordinate conversion lookup table 10
Is equal to the total number of image data to be cut out,
In the present invention, since the data compression step 22b has already been performed, the total number of data in the coordinate conversion lookup table 10 is significantly reduced. That is, when coordinate conversion is performed on an original image before compression as in the conventional method for identifying a circular pattern 36 shown in FIG. 6, the size is Wx × Wy as shown by a two-dot chain line in FIG. Lookup table 10 for coordinate conversion
However, in the case of the present invention, since coordinate conversion is performed on the compressed data, a coordinate conversion lookup table 10 having a size of Wx ′ × Wy ′ is sufficient as shown by a solid line in FIG. For example, in the data compression step 22b, 4 × 4
Assuming that a pixel is compressed to one pixel, the data compression ratio in the x direction is Cx = 4, the data compression ratio in the y direction is Cy = 4, and W
Since x ′ = Wx / Cx and Wy ′ = Wy / Cy, the size of the coordinate conversion lookup table 10 is Wx ′ × Wy ′ =
(Wx × Wy) / (Cx × Cy) = (Wx × Wy) / 1
It becomes 6. That is, the size of the coordinate conversion lookup table 10 is 1/16 that of the conventional method. For this reason, the capacity of the memory can be reduced, and the identification device can be downsized.
The cost can be reduced, and the amount of data handled in the coordinate transformation is reduced, so that the processing time can be reduced.

The amount of data handled in the feature emphasizing step 22a is larger than in the case of FIG. However, in the polar coordinate conversion step 22c, since the identification pattern 35 is formed by performing the coordinate conversion while referring to the position information of the coordinate conversion reference table 10 one by one, the data amount handled in the feature enhancement step 22a is increased. The minus plane is much smaller than the plus plane due to the decrease in the amount of data handled in the polar coordinate conversion step 22c. That is, even if the amount of data handled in the feature enhancement step 22a increases,
According to the present invention, the memory capacity can be reduced, and the identification device can be significantly reduced in size and cost.
Processing time can be greatly reduced.

Thereafter, the density conversion section 23 performs a density conversion step 23. In the density conversion step 23, density conversion processing (luminance adjustment processing) is performed, and the density conversion reference table 11 is used.
Is used to form a comparison pattern 38 from the identification pattern 35. This density conversion process associates each data of the identification pattern 35 with each data of the template 12 and replaces it with the data of the associated template 12, and is a process performed in the conventional pattern identification method. Although detailed description is omitted, for example,
No. 73438 "Concentration order table creation process"
And a “density conversion” composed of a “density conversion step”. In this case, the template density order table and the template density order table created from the template correspond to the density conversion reference table 11 of the present invention.

In the matching step 24, the dissimilarity or similarity between the comparison pattern 38 and the template 12 is calculated in the matching calculation unit 8. For example, the dissimilarity or similarity is calculated as follows. Now, FIG.
As shown in FIG.
Is x = 128 bits and y = 10 bits, and the (1,1) -th data of the comparison pattern 38 is represented by f
The (1,2) -th data is f2, ..., the (x, y) -th data is fi, ..., (128,10) -th data is f1280, and the (1,1) -th data of the template 12 is Data is t1, (2, 1) th data is t2,.
The (x, y) -th data is represented by ti,..., (128, 10)
The second data is set to t1280. Dissimilarity δ in this case
Is represented by Equation 1.

[0024]

(Equation 1)

The dissimilarity thus obtained is set to δ1. Then, the arrangement of each data of the comparison pattern 38 is shifted by one bit in the x direction, and the dissimilarity δ2 is similarly obtained. In this manner, the operation of shifting the arrangement of each data of the comparison pattern 38 by one bit in the x direction to obtain the dissimilarity δ is repeated 128 times to obtain the dissimilarities δ1 to δ128. Next, in the control / identification unit 9, the dissimilarities δ1 to δ
The smallest δmin among 128 is obtained, and it is determined whether or not δmin is smaller than a predetermined threshold th. For example, if a 100-yen coin is being identified, δ
When min <th, the degree of dissimilarity is small, so the discrimination target is determined to be a 100-yen coin, and when δmin ≧ th, the dissimilarity is large, so that the discrimination target is not a 100-yen coin. I do. Since the minimum value δmin is obtained from the dissimilarities δ1 to δ128 and the determination is made based on the minimum value δmin, the direction of the 100-yen coin when capturing an image of a 100-yen coin with the image input unit 1 ( rotation angle)
Is different from the direction when the template 12 was created, the determination can always be made in the direction that matches the direction when the template 12 was created. That is, δmin
In this case, it can be considered that the image of the 100-yen coin has been shot in the same direction as the direction in which the template 12 was created, and the dissimilarity can be determined from the relationship between δmin and the threshold th.

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 gist of the present invention. For example, in FIG. 2, the density conversion unit 7 is omitted, and the identification pattern 35 cut out by the image cutout unit 6 is compared with the template 12 using the comparison pattern 38 to calculate the similarity or the dissimilarity. good. In other words, the density conversion step 23 is omitted, and the identification pattern 35 formed in the polar coordinate conversion step 22c of the pre-processing step 22 is used as the comparison pattern 38 in the template 1 in the matching step 24.
You may make it compare with 2.

In FIG. 2, the image cutout section 6 is disposed immediately before the density conversion section 7, but may be disposed immediately after the density conversion section 7. That is, in FIG. 1, the data compression step 22b → the density conversion step 23 → the polar coordinate conversion step 22c may be performed. In this way, the density conversion is performed on the rectangular coordinates, so that the distortion of the characteristic information of the original image due to the polar coordinate conversion can be avoided, and the quality of the comparison pattern 38 can be prevented from deteriorating. .

Further, in the above description, the circular pattern 36
Is a ring shape, but may of course be a circular shape.

[0029]

As described above, in the method for identifying a circular pattern according to the first aspect, data compression is performed after performing a feature emphasizing step of enhancing the characteristics of the circular pattern, and the circular pattern after data compression is applied to the circular pattern. To perform coordinate transformation,
Since a comparison pattern was formed and the comparison pattern was compared with a reference template to identify a circular pattern, the pre-processing step was a feature enhancement step → a data compression step →
It is configured in the order of the polar coordinate conversion step, and the coordinate conversion can be performed on the data after compression. For this reason, the amount of data handled in the polar coordinate conversion process is reduced, so that the processing time can be shortened. In addition, the required memory capacity is reduced, and the identification device can be reduced in size and cost.

In this case, as in the method for identifying a circular pattern according to the second aspect, the circular pattern is obtained by reading an image to be identified and determining the density of each pixel as a circular ring in image information stored as a digital signal. It is desirable that the feature emphasizing process is a process of emphasizing a feature by differential processing. By doing so, it is possible to provide a pre-processing step of a method for identifying a circular pattern suitable for identifying a coin or the like.

According to a third aspect of the present invention, there is provided an apparatus for identifying a circular pattern, comprising: an image input unit for reading an image to be identified; an image memory for storing density of each pixel of the image to be identified read by the image input unit; A feature emphasis unit for emphasizing the density feature of the identification target image, a data compression unit for compressing the identification target image emphasized by the feature emphasis unit, and pattern identification from the identification target image compressed by the data compression unit. Since it is provided with an image cutout unit that cuts out a circular region to be performed, and a matching calculation unit that compares the region cutout by the image cutout unit with a pre-created reference template, An identification method can be implemented. That is, since the coordinate conversion is performed on the data after the compression, the amount of data to be handled when forming the identification pattern in the image cutout unit is reduced, so that the processing time can be shortened and the need is made. Since the memory capacity is reduced, the size and cost of the identification device can be reduced.

In this case, as in the circular pattern identification apparatus according to the fourth aspect, the image cutout section has a coordinate conversion reference table, and performs coordinate conversion on the circular area to generate an identification pattern having the same size as the template. It is desirable to cut it out. The template is a rectangular data array. Therefore, the image cutout unit uses the position information of the coordinate conversion reference table to coordinate-convert the image data of the circular area into a rectangular data array to form the identification pattern. Is subjected to coordinate transformation, and the processing time can be reduced, and the size and cost of the identification device can be reduced.

[Brief description of the drawings]

FIG. 1 is a flowchart illustrating a method for identifying a circular pattern according to the present invention.

FIG. 2 is a block diagram showing a schematic configuration of a circular pattern identification device of the present invention.

FIG. 3 shows a process of forming an identification pattern from an image to be identified in the method for identifying a circular pattern according to the present invention;
(A) is a conceptual diagram of a captured identification target image, (B) is a conceptual diagram of the identification target image after feature enhancement processing, (C) is a conceptual diagram of the identification target image after data compression, and (D) is polar coordinate transformation. It is a conceptual diagram of the identification pattern after.

FIG. 4 is a conceptual diagram of polar coordinate conversion in the method for identifying a circular pattern according to the present invention.

FIG. 5 is a conceptual diagram of a density conversion step and a matching step of the method for identifying a circular pattern according to the present invention.

FIG. 6 is a flowchart illustrating a conventional method for identifying a circular pattern.

7A and 7B show a process of forming an identification pattern from an image to be identified in a conventional method for identifying a circular pattern, and FIG.
Is a conceptual diagram of the captured identification target image, (B) is a conceptual diagram of the identification pattern after the polar coordinate conversion, (C) is a conceptual diagram of the identification pattern after the feature enhancement processing, and (D) is a conceptual diagram of the identification pattern after the data compression. It is a conceptual diagram.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Image input part 3 Image memory 4 Feature emphasis part 5 Data compression part 6 Image extraction part 8 Matching calculation part 10 Reference table for coordinate conversion 12 Template 22a Feature emphasis step 22b Data compression step 22c Polar coordinate conversion step 36 Circular pattern 37 Identification object Image 38 Comparison pattern

Claims (4)

[Claims] 1. A method for identifying a circular pattern, comprising: performing a coordinate transformation on a circular pattern to form a comparison pattern; and comparing the comparison pattern with a reference template to identify the circular pattern. After performing a feature emphasizing step of emphasizing the features of the above, data compression is performed, coordinate conversion is performed on the circular pattern after data compression, a comparison pattern is formed, and the comparison pattern is compared with a reference template. A method for identifying a circular pattern, characterized in that the circular pattern is identified. 2. The circular pattern is a circular ring-shaped pattern of image information in which an image to be identified is read and the density of each pixel is stored as a digital signal. 2. The method for identifying a circular pattern according to claim 1, wherein the process is an emphasis process. 3. An image input unit for reading an image to be identified, an image memory for storing the density of each pixel of the image to be identified read by the image input unit, and a feature of the density of the image to be identified is emphasized. A feature emphasizing unit to be coded, a data compressing unit for compressing the identification target image emphasized by the feature emphasizing unit, and an image obtained by cutting out a circular area for performing pattern identification from the identification target image compressed by the data compressing unit. An apparatus for identifying a circular pattern, comprising: a clipping unit; and a matching calculation unit that compares a region clipped by the image clipping unit with a previously created reference template. 4. The image cut-out unit has a coordinate conversion reference table, and performs coordinate conversion on the circular area to cut out an identification pattern having the same size as the template. An apparatus for identifying a circular pattern as described.