JP3007392B2 - Pattern recognition method and apparatus - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a pattern matching method for automatically detecting a binarized image to be inspected obtained by an image pickup apparatus in a non-contact manner, and an apparatus therefor. The present invention relates to a pattern recognition apparatus based on reference image pattern matching for performing pattern matching of an image to be inspected using a template to detect a position of a workpiece.

2. Description of the Related Art Template matching or pattern matching is performed by calculating the degree of approximation by counting the number of coincidences between corresponding pixels of a standard image to be compared with a partial image in an image to be inspected and searching for the most similar partial image. This is called matching processing. FIG. 12 is a principle diagram for explaining the outline of the pattern matching process, and is a premise of the present invention. Hereinafter, this will be described.

M rows × n columns (m ≦ M, n) in the inspection image 5 of M rows × N columns
≤ N), the pixel at the corresponding position of the reference image 7 for pattern matching of the same pixel size with the partial image 6 of ≤ N) is compared, and the number of matched pixels is counted. The number of comparison counts between the partial image 6 and the reference image 7 at a certain coordinate is m × n, and the number of input data is 2 × (m × n). As an example, when the size of the inspection image 5 is M = 320 pixels, N = 240 pixels, and the size of the reference image 7 is m = n = 64 pixels, the number of input data is 8192, and m = n.
In the case of = 128 pixels, the number of input data is 32654, and this amount of processing must be calculated as the similarity at one coordinate. The partial image 6 in the inspection image 5 is (M-m
+1) × (N−n + 1), and m × n comparisons must be performed for each of them, so that a total of m
× n × (M−m + 1) × (N−n + 1) comparison operations are required. That is, the size of the image 5 to be inspected is M = 32
When 0 pixels, N = 240 pixels, and the size of the reference image 7 is m = n = 64 pixels, an enormous amount of calculation is required, requiring 186,322,944 comparison operations.

For such a pattern matching process for automatically detecting a pattern of a two-dimensional image to be inspected in a non-contact manner at a high speed, for example, Japanese Patent Publication No. Sho 60-17152 and Japanese Patent Laid-Open No. Sho 62-210
No. 596 discloses an apparatus for performing an effective pattern matching process. The pattern matching system disclosed in Japanese Patent Publication No. Sho 60-17152 is generally called a composite partial pattern matching system, and its principle is that a two-dimensional pattern of a plurality of characteristic specific portions of an image to be inspected is used as a reference image 7.
The partial image 6 is sequentially cut out from the two-dimensional pattern of the image 5 to be inspected, and the cut-out partial image 6 is compared with the reference image 7 to detect a coordinate position that best matches the partial image 6. Things. Partial image 6
Is made to be 12 × 12 pixels. In order to perform recognition in real time by the feature of the pattern matching method, the calculation processing as the similarity at one coordinate must be completed within one pixel clock period, so that 12 × 12 = 144 pixels must be simultaneously performed. Therefore, the reference image data of 12 × 12 = 144 pixels must have a configuration capable of simultaneous output, and the inspection image data must also have a configuration capable of simultaneously outputting 144 pixels. Also, in order to complete the processing within one image scanning period, the comparison circuit for comparing the similarity must complete the arithmetic processing of the similarity at one coordinate within one pixel clock period. A logical OR circuit is required. Since these input / output data are similarities in one coordinate, 288 input wirings and 144 output wirings are required in an exclusive OR circuit to perform arithmetic processing at the same time.

In addition, the literature "Coarse-fine template matching" IEEE
Trans., SMC-7, pp. 104-107 (1977) and “Sequential
hierarchial scene matching '' IEEEETrans., C-27,4,35
According to page 9-366 (1978), pattern matching is not performed on an image (original image) with the highest image resolution, but an image with a coarser resolution is prepared in advance, and the image resolution is sequentially increased and the detection is performed. Hierarchical pattern matching methods for increasing efficiency are known. Conventionally, in a hierarchical pattern matching method or a coarse / fine detection method, which is a recognition method of performing pattern matching with a coarse image resolution and pattern matching on an image with a high image resolution, one pixel at a coarse resolution represents luminance data of a plurality of pixels. Gray-scale image processing that processes pixel data as coarse pixels based on the total value is described.Detection processing by this hierarchical pattern matching method is performed by software, and in order to perform real-time position detection Not yet.

As another method using the hierarchical pattern matching method, Japanese Patent Application Laid-Open No. Sho 62-55781 discloses that when pattern matching is performed at coarse pixel intervals, the image resolution is kept as it is, but the pattern matching is not performed in all areas of the reference pattern. By performing pattern matching only on partial areas, the processing speed is increased. In any of the above methods using the hierarchical pattern matching method, coarse pixels are not the reference image data itself registered in the original image, but the amount of reference image data is reduced during the comparison process by compressing or limiting the original image. We are trying to speed up processing.

A conventional apparatus for performing pattern matching using a reference pattern having an arbitrary shape and size is disclosed in
There is a method according to -210596. According to this method,
The reference pattern (8 × 12) is arranged at an arbitrary position to construct an n × m size reference pattern having an arbitrary size and an arbitrary shape. For each reference pattern in n × m,
The pattern matching is executed, and the result is shifted to the target pixel coordinates of the n × m reference pattern. The objective is achieved by adding the grayscale images obtained in this way after each shift.

Problems to be Solved by the Invention In the composite partial pattern matching method by providing a large number of small partial patterns, the size of the reference image 7 is small. Registration, which has the following problems.

1) In the pattern registration, since the size of the reference image 7 is small, it is necessary to perform the operation of registering a combination of characteristic partial patterns in order to perform the registration. there were.

2) In the pattern matching based on the part, since the size of the reference image 7 is small and the recognition rate is low in recognition of only one point, the pattern of three or four points is registered for one image area to be inspected. . For this reason, the relative position relationship (distance, angle) between each partial pattern must be checked in addition to the original method in which the position of the pattern can be detected only by the magnitude of the correlation value. Position cannot be detected.

3) When the same pattern larger than the size of the partial pattern exists in the pattern detection range of the image 5 to be inspected due to the small size of the partial pattern, the pattern position cannot be detected. For example, 3 in the pattern detection range
When there are two identical patterns, the position of the partial pattern which is registered as a partial pattern and matches the best is any one of the three coordinate positions as a candidate point when pattern matching is repeated, but cannot be uniquely determined.

4) Increasing the size of the partial pattern solves the above problem, but causes an inherent problem of pattern matching that the hardware scale is significantly increased.

5) Since the size of the partial pattern is about 12 × 12 pixels, it is easily affected by noise, and the position detection accuracy is not good.

Since noise is random in nature, the probability of occurrence of noise per pixel is generally uniform whether the size of the partial pattern is small or large. From this point, it is not directly related to the size of the partial pattern.
It is not only a pattern with a clear contrast that is easy to quantify. In this case, the noise is hard to be applied to a portion separated from the binarized threshold level because a video signal crossing the threshold level is hardly generated. However, a video signal level almost similar to the binarized threshold level is generated. Noise tends to occur on a binarized image in a pattern portion having a relatively large number as compared with other places. Therefore, there is a portion where the contrast is clear on the image to be inspected and a portion where the contrast is not clear. The pattern portion has a clear contrast and is hard to get noise when binarized, and the pattern is more characteristic than the pattern shape. Since the pattern part is independent of the pattern part, if it is registered as a partial pattern, it is easy to be affected by noise if the pattern part that is not clear contrast but is more characteristic than the pattern shape is registered. Gets worse. The effect can be reduced by increasing the number of partial patterns, but the processing time increases accordingly.

Japanese Patent Application Laid-Open No. Sho 62-5578, which is a hierarchical pattern matching method
No. 1 performs a pattern matching at a certain coarse pixel interval by creating a coarse reference pattern using a part of the original image limited by thinning out the reference image 7 at the time of coarse detection,
Whether the recognition is possible or not depends on the reference image 7. In addition, when registering the reference image 7, there is a problem in the ease of the registration work by selecting the quality of the worker.

In the method disclosed in Japanese Patent Application Laid-Open No. Sho 62-210596, the size of the reference image 7 is based on a small area of 8 × 12 pixels. For this purpose, processing must be performed using a large number of partial patterns, which requires processing time. Since the number of partial patterns to be registered changes depending on the pattern to be detected, the recognition time differs depending on the image to be inspected. The time management of the device equipped with the device is complicated. In addition, registering a large number of these partial patterns complicates the operation and judgment of the operator,
Registration was not easy.

Accordingly, the present invention has been proposed in view of the above-mentioned drawbacks, and has as its object to provide a new and improved pattern recognition device.

Means for Solving the Problems The following describes what technical means have been adopted to solve the above problems.

In a pattern recognition method for automatically extracting a specific pattern from a two-dimensional image, a two-dimensional video signal from an imaging unit is digitized, and an arbitrary size of a characteristic pattern of an image to be inspected is cut out and binarized in advance as a reference image. Reference image storage means capable of storing the image as it is as an original image and outputting image data stored at an arbitrary two-dimensional coordinate position, and an arbitrary number of dy horizontal scanning lines of the two-dimensional inspection image data set from a microcomputer. (VS-1) horizontal scanning line holding means capable of holding only the horizontal scanning lines are individually set to VS rows at k (1 ≦ k ≦ dy) horizontal scanning line intervals (= k pixel intervals). Vertical image data extracting means for simultaneously outputting the inspected image data corresponding to the inspection coordinate position; and inspecting image data from the vertical image data extracting means. By connecting the VS pieces simultaneous output of up to any dx pixels in the horizontal direction pixels only capable of holding the pixel shifting means is set from the microcomputer (HS-1) pieces series, 1
Set j with image data to be inspected in the horizontal direction at the same time
The image data to be inspected simultaneously output from the VS horizontal scanning lines from the vertical extraction means are simultaneously output from the HS locations evenly spaced at the pixel (1 ≦ j ≦ dx) interval. By providing the (HS-1) serial pixel shift means in the VS stage as input, the inspected image data corresponding to the individual inspection coordinate positions of the total VS × HS points of the two-dimensional inspected image data is simultaneously output. Horizontal image data extraction means, and the reference image data from the reference image storage means and the inspection image data corresponding to the individual inspection coordinate positions from the horizontal image data extraction means are simultaneously VS × HS points. Pattern comparison output corresponding to each inspection coordinate position of VS × HS point is output serially from each part corresponding to each inspection coordinate position as VS × HS coincidence data at the same time. And a coincidence data output corresponding to each inspection coordinate position from the logic operation means is simultaneously and individually counted to complete the pattern matching with VS × HS by one image scanning, and A portion corresponding to the inspection coordinate position is constituted by a correlation value accumulating means for independently calculating and holding a correlation value, and the first coarse detection is performed by using the horizontal image data extracting means and the vertical image data extracting means in the respective first pattern. Set so that HS × VS image data to be inspected is output at the same time at the pixel interval of collation.
In the correlation data of the HS × VS candidate points that were arranged so that the center of the inspection position of VS rows × HS columns corresponded, performed the first coarse detection, and performed the first pattern matching, Search for the part holding the maximum correlation value, if the value is greater than the predetermined coarse detection threshold value, the pattern matching coordinate position corresponding to the part holding the maximum correlation value, and the center coordinate position of the next pattern matching Is determined and arranged so that the center of the inspection position of VS rows × HS columns corresponds to the coordinate position of the maximum correlation value obtained in the first coarse detection, and the second coarse detection is performed horizontally. The direction is {first pixel interval−second pixel interval × (HS−1) / 2} ≦ 0, and the vertical direction is {first pixel interval−second pixel interval × (VS−1) / 2. ｝ ≦ 0, the pattern is arranged so that the center of the VS row × HS column corresponds to the second pixel interval at which the minimum integer value is satisfied. The second correlation is performed to determine the coordinate position of the highest correlation value of the correlation data of the second HS × VS candidate points, and the VS line is centered on the coordinate position of the highest correlation value obtained in the second coarse detection. × 3rd by arranging the center of the HS row
In the horizontal and vertical directions until the set pixel interval in the horizontal / vertical direction becomes one pixel interval, the horizontal direction is {pixel interval of the Pth pattern collation−pixel interval of the (P + 1) th pattern collation × (HS−1). / 2} ≦ 0, and the vertical direction is the smallest integer value that satisfies {pixel interval of P-th pattern matching−pixel interval of P + 1-th pattern matching × (VS−1) / 2} ≦ 0. At the pixel interval of the (P + 1) th pattern matching
The central portion of the VS row × HS column is arranged at the coordinate position of the Pth highest correlation value, and the maximum value of the correlation data of HS × VS candidate points of pattern matching performed at one pixel interval is predetermined. If the value is larger than the detection threshold value, the coordinate position corresponding to the portion holding the maximum correlation value is set as the detection position.

Another solution is a pattern recognition method for automatically extracting a specific pattern in a two-dimensional image, wherein the constant dx = 3
2, dy = 32, HS = 7, VS = 7, and the first coarse detection outputs image data to be inspected at an interval of 8 pixels for both the horizontal image data extraction means and the vertical image data extraction means. The first coarse detection is performed by arranging such that the center of the inspection position in the 7th row and 7th column (4th row and 4th column) corresponds to the center position of the detection range of the image to be inspected. If the maximum correlation value is larger than a predetermined coarse detection threshold value from among the correlation values of the coordinate positions of the 49th candidate point of the fourth round, the coordinate position corresponding to the part holding the maximum correlation value is determined, The second coarse detection is performed three times around the coordinate position obtained in the first coarse detection corresponding to the portion holding the maximum correlation value.
If the maximum correlation value is larger than a predetermined coarse detection threshold value among the correlation values of the coordinate positions of the 49 candidate points in the second time performed at the pixel interval, it corresponds to the portion holding the maximum correlation value The third detection is performed at one-pixel intervals around the coordinate position having the maximum correlation value obtained in the second coarse detection, and the coordinates of the third 49 candidate points are obtained. If the maximum correlation value among the position correlation values is larger than a predetermined detection threshold value, the detected position is a coordinate position corresponding to a portion holding the maximum correlation value, and the pattern is detected. Achieved by the method.

Still another solution is the solution according to item 1,
dx = 32, dy = 32, HS = 7, VS = 7, and the detection means is the first
At the 16th pixel interval, 7 × 7 = 49 pieces of image data to be inspected are set to be simultaneously output at the pixel interval of the first pattern matching, and 7 lines × 7 lines are set at the center position of the detection range of the image to be inspected.
The first coarse detection is performed by arranging the 7-column inspection positions so as to correspond to the center (4th row, 4th column), and a pattern matching position having a correlation value higher than the coarse detection threshold value cannot be obtained. In the second case, the pattern matching is performed at 16 pixel intervals in the same manner, but the matching position is calculated by performing recognition at a position where the image data generation position of the reference image pattern is shifted by 8 pixels. The detection threshold value is the same as the first coarse detection threshold value, and the second
If a candidate point larger than the second coarse detection threshold value is found, the correlation of the 49 second candidate points larger than the second coarse detection threshold value set in advance by the microcomputer. The coordinate position of the highest correlation value of the data is obtained, the third coarse detection is performed at three pixel intervals around the coordinate position of the highest correlation value obtained in the second coarse detection, and the microcomputer performs the second coarse detection. The coordinate position of the highest correlation value of the correlation data of the 49 candidate points for the third time, which has become larger than the third coarse detection threshold value set after the coarse detection, is obtained. The fourth and subsequent pattern matchings are performed at one-pixel intervals centering on the maximum value position coordinates obtained in the above, and the maximum value of the correlation data of the 49th candidate point in the fourth time is larger than a predetermined detection threshold value. If the value is It is achieved by the pattern recognition method which is characterized in that the detected position coordinate position corresponding to the holding portion a large correlation value.

Another solution is the solution described in item 1, wherein dx = 32, dy = 32, HS = 7, VS = 7, and the detection means performs the first pattern at 32 pixel intervals. 7 × 7 = 49 pieces of inspection image data are set to be output simultaneously at the pixel interval of the collation, and 7
The first coarse detection is performed by arranging the central part (4th row, 4th column) of the inspection position of the row × 7 columns so as to correspond, and the correlation value higher than the coarse detection threshold value also in the second pattern matching If the pattern matching position is not obtained, the pattern matching is performed in the second time using the first coarse detection threshold value at the same 32 pixel interval, but the position where the image data generation position of the reference image is shifted by 16 pixels is recognized. Is performed, the pattern matching position is calculated. If a pattern matching position having a correlation value higher than the coarse detection threshold value cannot be obtained, the pattern matching is performed for the third time at the same 32 pixel interval, but the reference image The recognition position is calculated by recognizing the image data generation position at a position shifted by 8 pixels from the second time, and the third coarse detection threshold value is set to the first time. A coordinate position having the maximum correlation value is obtained from among the correlation values of the coordinate positions of the 49 candidate points of the third time, which is the same value as the detection threshold value and is larger than the coarse detection threshold value for the first time, The fourth coarse detection is performed at three pixel intervals centering on the coordinate position obtained in the third coarse detection, and the maximum correlation value among the correlation values of the coordinate positions of the 49th candidate point in the fourth time The fifth detection is performed at one-pixel intervals centering on the coordinate position obtained by the fourth coarse detection, and the fifth correlation value of the 49 candidate points is obtained. If the maximum correlation value is larger than a predetermined detection threshold value, the detection position is obtained by using a coordinate position corresponding to the portion holding the maximum correlation value. Is done.

Still another solution is the solution according to item 1, wherein
Interpolation processing using the HS × VS pattern matching correlation data and inspection coordinate positions performed at one pixel intervals makes horizontal and horizontal
According to a pattern recognition method, a collation coordinate position of one pixel or less in both vertical directions is calculated and set as a detection position.
In addition, the pattern matching position is horizontally determined by using the correlation data of the pattern matching performed at one pixel interval. From the spline interpolation calculation using a total of three pieces of correlation data of the upper and lower adjacent coordinates including the coordinate position having the highest correlation value, 1 in both the horizontal and vertical directions
Each is achieved by a pattern recognition method characterized in that a collation coordinate position of a pixel or less is calculated and set as a detection position.

The above-described recognition method can further detect the pattern matching state at the candidate point at high speed by simultaneously comparing and determining the correlation value data of the correlation value accumulating means with a pattern detection threshold value set in advance by a microcomputer. Then, the data output of the correlation value accumulating means is compared with the held detection threshold value, and the magnitudes are simultaneously compared and determined between the coordinate positions corresponding to the individual inspection coordinate positions, and the individual inspections holding the results are performed. The holding means corresponding to the coordinate position is provided for each of the VS × HS individual inspection coordinate positions, and the microcomputer collects the comparison results of one horizontal scanning line HS and collects the comparison results at one time by the microcomputer. The means that holds the state so that the state of each pattern matching position of one horizontal scanning line can be grasped by reading By providing only the book, a horizontal candidate point detecting means is provided which is capable of quickly grasping the outline of the state of each pattern matching position of each horizontal scanning line and specifying the position corresponding to the calculation, and In the coarse detection of the second time, the horizontal image data extraction means and the vertical image data extraction means are set to HS × VS at the pixel interval of the first pattern matching.
Is set so that the image data to be inspected are output at the same time, and the center of the inspection position of VS rows × HS columns is arranged to correspond to the center position of the detection range of the image to be inspected. After the first coarse detection by the horizontal candidate point detecting means, the microcomputer checks whether or not there is a candidate point at a detection position exceeding the first coarse detection threshold value set in advance by the microcomputer. The information of the horizontal candidate point detection means of the horizontal comparison line indicating the existence of the first comparison pattern is read from the means for holding the correlation value at the pattern matching coordinate position corresponding to the information. The pattern coordinate position corresponding to the part holding the maximum correlation value among the correlation count values of the plurality of candidate points subjected to the collation is determined as the center coordinate position of the next pattern collation. Are arranged so that the center of the inspection position of the HS row × VS column corresponds to the center of the coordinate position of the highest correlation value obtained by the coarse detection, and the second coarse detection is performed in the horizontal direction of the first pixel. Interval−second pixel interval × (HS−1) / 2｝ ≦
0, the vertical direction is {first pixel interval−second pixel interval ×
The (VS-1) / 2｝ ≦ 0 is arranged so that the center of the VS row × HS column corresponds at the second pixel interval that satisfies (VS-1) / 2｝ ≦ 0. The coordinate position of the highest correlation value of the correlation data of the plurality of candidate points for the second time, which is larger than the second coarse detection threshold value set after the coarse detection, is determined. The horizontal direction is the pixel interval of the P-th pattern matching until the third and subsequent pattern matchings are performed at one-pixel intervals by arranging the center of the VS row and HS column with the maximum value position coordinates obtained as the center. -Pixel interval of the (P + 1) th pattern matching x
(HS-1) / 2} ≦ 0, and the vertical direction satisfies {pixel interval of P-th pattern matching−pixel interval of P + 1-th pattern matching × (VS−1) / 2} ≦ 0. The center of the VS row ≦ × HS column is arranged at the pixel interval of the (P + 1) th pattern matching, and the pattern matching is performed at one pixel interval by the microcomputer from the horizontal candidate point detecting means.
P + 1 that has become larger than the detection threshold value for the first time
This is achieved by a pattern recognition method characterized by using the coordinate position having the maximum value of the correlation data of the single or plural candidate points at the time of the detection as the detection position.

Still another second solution is a pattern recognition method for automatically extracting a specific pattern from a two-dimensional image, comprising: digitizing a two-dimensional video signal from an imaging means; A reference image pattern storage means capable of outputting image data previously stored in a memory as a cut-out reference image, and capable of holding an arbitrary number of horizontal scanning dy lines of the two-dimensional image data to be inspected by an amount set by the microcomputer (VS-1) horizontal scanning line holding means are set to (1 ≦ k ≦
dy) vertical image data extraction means for simultaneously outputting inspection image data corresponding to individual inspection coordinate positions of VS rows evenly spaced at horizontal scanning line intervals (k pixel intervals); (HS-1) serially connected pixel shift means capable of holding VS simultaneous output of image data from the image data extracting means up to an arbitrary number of pixels dx in the horizontal direction by the number of pixels set by the microcomputer. As a result, image data to be inspected in one horizontal direction are evenly spaced at a set pixel (1 ≦ j ≦ dx) interval (j pixel interval) at the same time.
The image data to be simultaneously output from the HS locations is input to the image data to be inspected simultaneously output from the VS horizontal scanning lines from the vertical extracting means, and the (HS-1) serial pixel shifting means is used. A horizontal image data extracting unit configured to simultaneously output the inspection image data corresponding to the individual inspection coordinate positions of the total VS × HS points of the two-dimensional inspection image data by providing a VS stage; and the reference image storage unit. A pattern corresponding to each inspection coordinate position consisting of VS × HS exclusive ORs for simultaneously comparing patterns of image data from the image data to be inspected corresponding to individual inspection coordinate positions from the horizontal image data extracting means. A logical operation means for outputting a collation output as VS × HS simultaneous coincidence pulses from exclusive OR means corresponding to individual inspection coordinate positions; and By simultaneously counting each matched pulse output corresponding to each inspection coordinate position, VS x HS pattern matching is completed in one image scan, and the correlation is calculated and held independently in the portion corresponding to each inspection coordinate position A value accumulating means, and a pattern matching state at a candidate point can be detected at high speed by simultaneously comparing and judging the correlation value data of the correlation value accumulating means with a pattern detection threshold value set in advance by a microcomputer. Comparing the data output of the correlation value accumulating means with the held detection threshold value, and comparing and judging the magnitudes of the coordinate positions corresponding to the individual inspection coordinate positions at the same time; The holding means corresponding to the position is provided for each of the VS × HS individual inspection coordinate positions, and the microcomputer has one vertical scanning line. A means for keeping the state of each pattern matching position of one vertical scan line by one reading of the microcomputer by combining the comparison results of VS locations is provided only for HS lines that perform pattern matching in the horizontal direction. In this way, the first rough detection is made up of horizontal image data extraction means and vertical image data extraction Both means are set so that image data to be inspected are simultaneously output at the first pixel interval, and are arranged such that the center of the inspection position of VS rows × HS columns corresponds to the center position of the detection range of the image to be inspected. The first coarse detection is performed, and the first coarse detection threshold value preset by the microcomputer after the first coarse detection by the vertical candidate point detecting means. A check is made to see if there is a candidate point at the detection position exceeding, and if so, the information of the vertical candidate point detecting means of the vertical comparison line indicating the existence is read and the correlation at the pattern matching coordinate position corresponding to the information is read. The pattern holding coordinate position corresponding to the part holding the maximum correlation value among the correlation count values of the plurality of candidate points for which the first pattern matching has been performed is determined by the means holding the value. The center coordinate position is determined, and the coordinate position of the highest correlation value obtained in the first coarse detection is set to VS.
Arrange so that the center of the inspection position of row x HS column corresponds,
The second coarse detection is performed in the horizontal direction by {1st pixel interval−2nd pixel interval}.
Pixel interval × (HS-1) / 2} ≦ 0, and the vertical direction is {1st
Pixel interval−second pixel interval × (VS−1) / 2｝ ≦ 0,
At the second pixel interval to be filled, VS rows ≦ HS columns are arranged so as to correspond to each other, and the second candidate set by the microcomputer from the vertical candidate point detecting means after the first coarse detection by the microcomputer. The coordinate position of the highest correlation value of the correlation data of the plurality of candidate points in the second time that has become larger than the coarse detection threshold value of the second detection is calculated, and the VS line is centered on the highest value position coordinates obtained in the second coarse detection. The horizontal direction is ま で the pixel interval of the Pth pattern matching−the pixel interval of the (P + 1) th pattern matching × ( HS-1) / 2} ≦ 0, and the vertical direction is {Pth pixel interval of pattern matching−P +
Pixel interval of first pattern matching × (VS−1) / 2｝ ≦ 0
At the pixel interval of the (P + 1) th pattern matching that satisfies
The central part of VS rows × HS columns is arranged and the pattern matching is performed by the microcomputer from the vertical candidate point detecting means.
A pattern characterized in that a coordinate position having a maximum value of correlation data of single or plural candidate points of the (P + 1) th time that is larger than a detection threshold value for the (P + 1) th time performed at a pixel interval is set as a detection position. Achieved by recognition methods.

Still another solution is a pattern recognition method for automatically extracting a specific pattern from a two-dimensional image, wherein a two-dimensional video signal from an imaging means is digitized, and an arbitrary size of a characteristic pattern of an image to be inspected is cut out. A reference image storage means capable of outputting image data previously stored in a memory as a reference image; and an arbitrary number of dy horizontal scanning lines of the two-dimensional image data to be inspected, which can be held by an amount set by the microcomputer (VS-1 ) Horizontal scanning line holding means (1 ≦ k ≦
dy) vertical image data extraction means for simultaneously outputting inspection image data corresponding to individual inspection coordinate positions of VS rows evenly spaced at horizontal scanning line intervals (k pixel intervals); A series of (HS-1) pixel shift means capable of holding VS simultaneous output of the image data to be inspected from the image data extracting means up to an arbitrary number of dx pixels in the horizontal direction by the number of pixels set by the microcomputer. By connecting to
The image data to be inspected in one horizontal direction is simultaneously output from the HS points evenly spaced at a set pixel j (1 ≦ j ≦ dx) interval at the same time. With the image data to be inspected simultaneously output from the plurality of horizontal scanning lines as inputs, the total number VS of the two-dimensional image data to be inspected can be obtained by providing the HS serial pixel shift means in the VS stage.
A horizontal image data extracting means for simultaneously outputting the inspected image data corresponding to each of the inspection coordinate positions of the × HS point, and image data from the reference image storing means and individual horizontal image data from the horizontal image data extracting means. VS for pattern comparison of inspected image pattern data corresponding to the inspection coordinate position
A logic consisting of outputting pattern matching outputs corresponding to individual inspection coordinate positions consisting of × HS exclusive ORs from the exclusive OR means corresponding to individual inspection coordinate positions as VS × HS simultaneous coincidence pulses The calculating means and the coincidence pulse output corresponding to each inspection coordinate position from the logical operation means are individually and simultaneously counted, whereby VS × HS pattern matching is completed in one image scan, and each inspection coordinate position is obtained. Correlation value integrating means for independently calculating and holding correlation values in corresponding portions, and simultaneously comparing and determining the correlation value data of the correlation value integrating means with a pattern detection threshold value preset by a microcomputer. The data output of the correlation value accumulating means is compared with the stored detection threshold value so that the pattern matching state at the candidate point can be detected at a high speed, and the individual At the same time, the magnitude is compared and determined between the coordinate position corresponding to the inspection coordinate position and the holding means corresponding to each inspection coordinate position holding the result is provided by the amount corresponding to VS × HS individual inspection coordinate positions. The microcomputer is designed to collect the comparison results of one horizontal scanning line HS and collect the comparison results of one horizontal scanning line at a time so that the outline of the state of each pattern matching position of one horizontal scanning line can be quickly grasped by one reading of the microcomputer. By providing only VS means for vertical pattern matching, a horizontal candidate that can quickly grasp the outline of the state of each pattern matching position of each horizontal scanning line and easily specify the position corresponding to the calculation. A point detection means, and a microcontroller for enabling a quick overview of the state of each pattern matching position of the VS horizontal correlation lines of the horizontal candidate point detection means. The output of each state holding means corresponding to each of the VS individual inspection coordinate positions of the horizontal candidate point detection means in each state holding means corresponding to each inspection coordinate position so that the data can be grasped by one data input by the computer. If at least one state is held larger than the coarse detection threshold value, the state of the horizontal correlation line is held to notify the microcomputer that there is a horizontal correlation line exceeding the coarse detection threshold value. In the first coarse detection, the horizontal image data extracting means and the vertical image data extracting means simultaneously output HS × VS inspection image data at the pixel interval of the first pattern matching. To the center of the detection range of the image to be inspected.
The first coarse detection is performed by arranging the inspection positions of the HS rows × VS columns so as to correspond to each other, and is set in advance by the microcomputer after the first coarse detection by the candidate point detecting means. It checks whether there is a candidate point at the detection position exceeding the first coarse detection threshold value, and if so, reads the information of the horizontal candidate point detecting means of the horizontal comparison line indicating the existence, and reads the information. By means of holding the correlation value at the pattern matching coordinate position corresponding to
A pattern coordinate position corresponding to a portion holding the maximum correlation value among the correlation count values of a plurality of candidate points subjected to the first pattern matching is determined as a center coordinate position for the next pattern matching, and The inspection is arranged so that the center of the inspection position of HS rows × VS columns corresponds to the center of the coordinate position of the highest correlation value obtained in the first coarse detection, Pixel interval−second pixel interval × (HS−
1) / 2} ≦ 0, and HS rows × VS columns at a second pixel interval that satisfies {first pixel interval−second pixel interval × (VS−1) / 2} ≦ 0 in the vertical direction. Are arranged so that the central portions thereof correspond to each other, and the candidate point detecting means sets a second plurality of second detection values larger than a second detection threshold value set by the microcomputer after the first detection. The coordinate position of the highest correlation value of the correlation data of the candidate points is determined, and HS rows × VS are centered on the highest value position coordinates obtained in the second coarse detection.
Arrange the center of the column to perform pattern matching for the third and subsequent
In the horizontal direction, the pixel interval of the P-th pattern matching−the pixel interval of the (P + 1) -th pattern matching × (HS−1) / 2} ≦ 0 in the horizontal direction and the {P-th The center of HS rows × VS columns is arranged at the pixel interval of the (P + 1) th pattern matching that satisfies the pixel interval of the pattern matching−the pixel interval of the (P + 1) th pattern matching × (VS−1) / 2｝ ≦ 0. And the microcomputer performs pattern matching at one pixel interval from the horizontal candidate point detection means.
The (P +) th that has become larger than the detection threshold value for the + 1st time
This is achieved by a pattern recognition method characterized in that a coordinate position having the maximum value of correlation data of single or multiple candidate points for the first time is set as a detection position.

Still another fourth solution is a pattern recognition method for automatically extracting a specific pattern from a two-dimensional image, comprising: digitizing a two-dimensional video signal from an imaging device; A reference image pattern storage means capable of outputting image data previously stored in a memory as a cut-out reference image pattern; and an arbitrary number of horizontal scanning lines dy of two-dimensional inspection image data
(VS-1) horizontal scan line holding means (1 ≦ k ≦
dy) vertical image data extraction means for simultaneously outputting inspection image data corresponding to individual inspection coordinate positions of VS rows evenly spaced at horizontal scanning line intervals (k pixel intervals); A series of (HS-1) pixel shift means capable of holding VS simultaneous output of the image data to be inspected from the image data extracting means up to an arbitrary number of dx pixels in the horizontal direction by the number of pixels set by the microcomputer. By connecting to
The image data to be inspected in one horizontal direction are simultaneously output from HS locations evenly spaced at a set pixel j (1 ≦ j ≦ dx) interval (j pixel interval) at the same time. The image data to be inspected simultaneously output from the VS horizontal scanning lines from the extraction means is input to the (HS-
1) Horizontal image data extraction in which the inspection image data corresponding to each inspection coordinate position of the total VS × HS point of the two-dimensional inspection image data is simultaneously output by providing the series pixel shift means in the VS stage. Means for simultaneously comparing the image data from the reference image storage means and the inspected image pattern data corresponding to the individual inspection coordinate positions from the horizontal image data extraction means.
A logic consisting of outputting pattern matching outputs corresponding to individual inspection coordinate positions consisting of × HS exclusive ORs from the exclusive OR means corresponding to individual inspection coordinate positions as VS × HS simultaneous coincidence pulses The calculating means and the coincidence pulse output corresponding to each inspection coordinate position from the logical operation means are individually and simultaneously counted, whereby VS × HS pattern matching is completed in one image scan, and each inspection coordinate position is obtained. Correlation value accumulating means for independently calculating and holding in the corresponding portion, and simultaneously comparing and judging the correlation value data of the correlation value accumulating means with a pattern detection threshold value set in advance by a microcomputer at a candidate point. In order to detect the pattern matching state at a high speed, the data output of the correlation value integrating means is compared with the held detection threshold value, and each inspection coordinate is compared. The size is compared and judged simultaneously with the coordinate position corresponding to the position, and the holding means corresponding to the individual inspection coordinate position to hold the result is provided only for VS × HS individual inspection coordinate positions. The microcomputer collects the comparison results of one vertical scanning line VS location and keeps the state so that the outline of the state of each pattern matching position of one vertical scanning line can be quickly grasped by one reading of the microcomputer. By providing only the HS lines that perform pattern matching in the horizontal direction, the above-mentioned means can be used to quickly grasp the outline of the state of each one vertical scanning line and to easily specify the position corresponding to the operation, and a vertical candidate point detecting means. In order to grasp at a high speed an outline of the state of each pattern matching position of the HS vertical correlation lines of the vertical candidate point detecting means, The outputs of the state holding means corresponding to the individual inspection coordinate positions of the HS lines of the vertical candidate point detection means can be grasped by a single data input. Candidate point detection means for notifying the microcomputer that there is a vertical correlation line exceeding the coarse detection threshold value by holding the state of the vertical correlation line if at least one state is held larger than the coarse detection threshold value. In the first coarse detection, the horizontal image data extraction unit and the vertical image data extraction unit are configured to simultaneously output HS × VS image data to be inspected at the pixel interval of the first pattern matching. To the center of the detection range of the inspection image.
The first coarse detection is performed by arranging the inspection positions of the HS rows × VS columns so as to correspond to each other, and is set in advance by the microcomputer after the first coarse detection by the vertical candidate point detecting means. It checks whether there is a candidate point at the detection position exceeding the first coarse detection threshold value, and reads the information of the vertical candidate point detection means of the vertical comparison line indicating the existence when the existence exists. The means for holding the correlation value at the pattern matching coordinate position corresponding to the information is the pattern corresponding to the part holding the maximum correlation value among the correlation count values of the plurality of candidate points for which the first pattern matching has been performed. The collation coordinate position is determined as the central coordinate position of the next pattern collation, and the center of the inspection position of the HS row x VS column corresponds to the coordinate position of the highest correlation value obtained in the first coarse detection. Do Uni arranged, the horizontal direction of the second round of coarse detection {first pixel interval - the second pixel interval × (HS
-1) / 2} ≦ 0, and the vertical direction is {first pixel interval−second
Pixel spacing × (VS−1) / 2｝ ≦ 0, and the second pixel spacing is satisfied so that the center of the HS row × VS column corresponds to each other. The coordinate position of the highest correlation value of the correlation data of the second plurality of candidate points that has become larger than the second coarse detection threshold value set after the first coarse detection is obtained.
VS line x centering on the highest position coordinate obtained in the second coarse detection
Until the third and subsequent pattern matchings are performed at one-pixel intervals with the center portion of the HS row being arranged, the horizontal direction is {pixel interval of the P-th pattern matching−pixel interval of the (P + 1) -th pattern matching × (HS -1) / 2} ≦ 0, and in the vertical direction, the (P + 1) th time that satisfies {P-th pixel interval of pattern matching−P + 1-th pixel interval of pattern matching × (VS−1) / 2} ≦ 0. The detection threshold value for the (P + 1) th time, in which the center of VS rows × HS columns is arranged at the pixel interval of the pattern matching, and the pattern matching is performed at one pixel interval by the microcomputer from the vertical candidate point detecting means. No. P which became bigger
This is achieved by a pattern recognition method characterized in that a coordinate position having the maximum value of correlation data of a single or a plurality of candidate points at the (+1) th time is set as a detection position.

Further, in the above-described pattern recognition method having the first solving means, in the first coarse detection, the image data to be inspected is output at the first pixel interval in both the horizontal image data extracting means and the vertical image data extracting means. The first coarse detection is performed by arranging such that the center of the inspection position of VS rows × HS columns corresponds to the center position of the detection range of the inspection image, and performs the first coarse detection. When a candidate point whose output is larger than the coarse detection threshold value for the first time preset by the microcomputer is found, the horizontal candidate point detecting means detects the candidate point for the first time by the microcomputer. The position of the center of gravity is determined from the XY coordinate position data indicating the presence or absence of the single or plural pattern matching candidate points that have become larger than the coarse detection threshold value of the first time. The second coarse detection is performed at the second pixel interval that satisfies {first pixel interval−second pixel interval × (HS−1) / 2} ≦ 0 centering on the center of gravity obtained in the above. First from the microcomputer from the candidate point detection means
The position of the center of gravity of the position data of a single or a plurality of candidate points for the second time, which is larger than the second coarse detection threshold value set after the second coarse detection, is determined. Until the third and subsequent pattern matchings are performed at one pixel intervals centering on the position of the center of gravity obtained as described above, {the pixel interval of the Pth pattern matching−the pixel interval of the (P + 1) th pattern matching × (HS−1) / It is performed at the pixel interval of the (P + 1) -th pattern matching that satisfies 2｝ ≦ 0, and is larger than the (P + 1) -th detection threshold value set after the second coarse detection by the microcomputer from the candidate point detecting means. This is achieved by a pattern recognition method characterized in that the detected position has the maximum value of the position data of the (P + 1) th single or multiple candidate points.

In the above-described pattern recognition method having the second solving means, the first coarse detection is performed by the horizontal image data extracting means.
Both the vertical image data extracting means are set so that the image data to be inspected is output at the first pixel interval, and the center of the inspection position of HS rows × VS columns corresponds to the center position of the detection range of the image to be inspected. When the first coarse detection is performed by arranging, and a candidate point whose output from the vertical candidate point high-speed detecting means is larger than a coarse detection threshold value for the first time preset by a microcomputer is found. XY coordinate position data indicating the presence or absence of a first or a plurality of pattern matching candidate points larger than a first coarse detection threshold value set in advance by the microcomputer from the vertical candidate point detecting means. The position of the center of gravity is further determined, and the second coarse detection is performed based on the position of the center of gravity determined in the first coarse detection, which is {(first pixel interval−second pixel interval × (HS
-1) Performed at a second pixel interval that satisfies / 2｝ ≦ 0, and becomes larger than the second coarse detection threshold value set after the first coarse detection by the microcomputer from the candidate point detecting means. The second time the center of gravity of the position data of single or multiple candidate points is obtained, and the third and subsequent pattern matchings are performed at one pixel intervals with the center of gravity obtained by the second coarse detection as the center. Until this is done, the pixel interval of the (P + 1) th pattern matching that satisfies {P pixel interval of the Pth pattern matching−pixel interval of the (P + 1) th pattern matching × (HS−1) / 2} ≦ 0 is satisfied. The point detecting means has the maximum value of the position data of the (P + 1) th single or plural candidate points which is larger than the (P + 1) th detection threshold value set after the second coarse detection by the microcomputer. Inspection It is achieved by the pattern recognition method which is characterized in that the position.

Here, in the first solution, dx = 32, dy = 32, HS =
7, VS = 7, and the detection means is the first coarse detection of 8 for both the horizontal image data extraction means and the vertical image data extraction means.
The first coarse detection is performed by setting the image data to be inspected so as to be output at pixel intervals, and arranging the central position of the inspection position in 7 rows and 7 columns corresponding to the central position of the detection range of the image to be inspected. When a candidate point whose output from the horizontal candidate point high-speed detection means is larger than a first coarse detection threshold value set in advance by a microcomputer is found, the horizontal candidate point high-speed detection means The position of the center of gravity is determined based on the correlation value of the position data of the single or multiple candidate points of the first time which is larger than the first coarse detection threshold value for the first time set by the microcomputer. The second coarse detection is performed at three pixel intervals centering on the center of gravity determined in the second coarse detection, and is set by the microcomputer from the candidate point high-speed detection means after the first coarse detection. The barycentric position of the position data of the second or single candidate point which has become larger than the second coarse detection threshold value is determined, and the third barycenter position is determined based on the barycentric position obtained in the second coarse detection. The third detection is performed at one pixel interval from the candidate point high-speed detection means and becomes larger than the third coarse detection threshold value set after the second coarse detection by the microcomputer from the microcomputer. Alternatively, this is achieved by a pattern recognition method characterized in that the detected position has the maximum value of the position data of a plurality of candidate points.

In addition, the present invention is achieved by a pattern recognition method characterized in that when there is no coordinate position in a range of an image to be inspected that reaches a predetermined pattern matching correlation value, detection of a target position is disabled.

Still another solution is to provide a binarization circuit for digitizing a two-dimensional video signal from the imaging device, and a memory in which a characteristic pattern of an image to be inspected is cut out as an arbitrary size within a storage capacity range of the memory and a reference image is previously stored A reference image storage circuit capable of outputting the image data stored in the storage unit; and optionally converting the binarized inspection image data from the binarization circuit from one horizontal scanning line to a dx horizontal scanning line according to an instruction from the microcomputer. VS-1 horizontal scanning line shift registers (VS-1) capable of holding only the amount set in (1), and binarized by the (VS-1) horizontal scanning line shift registers and the binarizing circuit. Inspection image data corresponding to individual inspection coordinate positions of VS rows evenly spaced at k horizontal scanning lines (1 ≦ k ≦ dx) intervals (= k pixel intervals) set in the vertical direction of the image are simultaneously displayed. A vertical image data extraction circuit that is to be output; and VS simultaneous outputs of the binarized image data to be inspected from the vertical image data extraction circuit, which are output from one pixel to d pixels in the horizontal direction from a microcomputer. By connecting HS variable-length pixels and shift registers that can hold the pixels arbitrarily set by the instruction of 直列 in series, the set j pixel (1) in which binary image data to be inspected in one horizontal direction is simultaneously present 1 ≦ j ≦ dy) are simultaneously output from the HS locations evenly opened at intervals, and
The binary serial inspection image data output simultaneously from the VS horizontal scanning lines is input to the HS serial variable length pixels.
By providing a shift register in the VS stage, the total of the binary image can be calculated.
A horizontal image data extraction circuit configured to simultaneously output binarized image data to be inspected corresponding to each inspection coordinate position of a VSxHS point; reference image data from the reference image storage circuit; and the horizontal image data extraction circuit VS × HS for pattern comparison of inspected image data corresponding to each inspection coordinate position from
A logical operation consisting of outputting pattern matching outputs corresponding to individual inspection coordinate positions composed of exclusive OR circuits as VS × HS coincident pulses from the exclusive OR unit corresponding to individual inspection coordinate positions at the same time The circuit and the coincidence pulse output corresponding to each inspection coordinate position from the logical operation circuit are individually and simultaneously counted, whereby VS × HS pattern matching is completed in one image scanning period and each inspection coordinate position is obtained. Correlation value accumulating circuits that calculate and hold the corresponding parts independently are provided. In the first coarse detection, the image data to be inspected is simultaneously output at the first pixel interval in both the horizontal image data extraction unit and the vertical image data extraction unit. The first coarse detection is performed by arranging such that the center of the inspection position of VS row × HS column corresponds to the center position of the detection range of the inspection image, and the first coarse detection is performed.
The coordinate position of the highest correlation value of the correlation data of the VS × HS candidate points is obtained, and the second coarse detection is performed in the horizontal direction around the coordinate position of the highest correlation value obtained in the first coarse detection. ｛First
Pixel interval−second pixel interval × (HS−1) / 2｝ ≦ 0,
The vertical direction is {first pixel interval−second pixel interval × (VS−
1) / 2｝ ≦ 0 at the second pixel interval that satisfies the minimum integer value, and obtains the coordinate position of the highest correlation value of the correlation data of the second VS × HS candidate points, The third and subsequent pattern collations are performed in the horizontal direction until the set pixel interval in the horizontal and vertical directions becomes one pixel interval with the position coordinate having the highest correlation value obtained in the second coarse detection as the center. The pixel interval of the pattern matching−the pixel interval of the (P + 1) th pattern matching × (HS−1) / 2} ≦ 0, and the vertical direction is {the pixel interval of the Pth pattern matching—the pixel of the (P + 1) th pattern matching The interval × (VS−1) / 2｝ ≦ 0 is satisfied at the pixel interval of the (P + 1) th pattern matching that satisfies the minimum integer value, and HS × VS candidates for pattern matching performed at one pixel interval If the maximum value of the correlation data at the point is a value larger than a predetermined detection threshold value, It is achieved by a pattern recognition device which is characterized in that the detected position coordinate position corresponding to the holding portion a large correlation value.

On the other hand, in a pattern recognition device for automatically extracting a specific pattern from a two-dimensional image, a binarizing circuit for digitizing a two-dimensional video signal from an imaging device, and storing a characteristic pattern of an image to be inspected in a memory. A reference image storage circuit capable of outputting image data stored in an arbitrary two-dimensional coordinate position in advance as a clipped reference image as an arbitrary size within a capacity range, and a binarized inspection image from the binarization circuit It is composed of horizontal scanning line shift registers (VS-1) capable of holding data from one horizontal scanning line to dx horizontal scanning lines arbitrarily set by an instruction from the microcomputer. K horizontal scanning lines (1 ≦ k ≦ dx) intervals (= k pixels) set in the vertical direction of the binarized image by the horizontal scanning line shift registers and the binarizing circuit. A vertical image data extraction circuit configured to simultaneously output inspection image data corresponding to individual inspection coordinate positions of a VS row that is evenly opened at intervals, and a binarized image from the vertical image data extraction circuit. HS variable-length pixels and shift registers that can hold the VS simultaneous output of the inspected image data from 1 pixel to d pixel in the horizontal direction by the number of pixels set arbitrarily by the instruction from the microcomputer are connected in series. By connecting the pixels, the image data to be inspected which has been binarized in one horizontal direction is simultaneously set at a certain pixel (1
.Ltoreq.j.ltoreq.dy, which are simultaneously output from HS locations evenly spaced at intervals (j pixel intervals), and are binarized simultaneously output from VS horizontal scanning lines from the vertical extraction circuit. By providing the VS stages of the HS serial variable length pixels and shift registers with the image data to be inspected as input, the binarized image data to be inspected corresponding to each inspection coordinate position of the total VSxHS point of the binarized image can be obtained. A horizontal image data extraction circuit that is simultaneously output; and image data from the reference image storage circuit and pattern image data to be inspected corresponding to each inspection coordinate position from the horizontal image data extraction circuit are simultaneously subjected to pattern comparison. VS
The pattern matching output corresponding to each inspection coordinate position consisting of × HS exclusive OR circuits is simultaneously output as VS × HS matching pulses from the exclusive OR unit corresponding to each inspection coordinate position. A logic operation circuit, VS × HS pattern matching is completed in one field by simultaneously counting each coincidence pulse output corresponding to each inspection coordinate position from the logical operation circuit, and the individual inspection coordinate positions are obtained. Correlation value integration circuits that independently calculate and hold in the corresponding portions, and simultaneously compare and determine the correlation value data of the correlation value integration circuit with a pattern detection threshold value set in advance by a microcomputer at candidate points. The output of the correlation value integrating circuit is provided with a comparator corresponding to each inspection coordinate position so that a pattern matching state can be detected at high speed. The latches corresponding to the individual inspection coordinate positions that hold and hold the results are compared and judged simultaneously with the latches corresponding to the individual inspection coordinate positions, and the results are retained. One horizontal scanning line is provided for the microcomputer.
By providing only VS latches for vertical pattern collation, the comparison results of HS locations are combined and the state of each pattern collation position of one horizontal scan line can be grasped by one reading of the microcomputer. A horizontal candidate point detection circuit capable of grasping the state of each pattern matching position of the scanning line and specifying the position corresponding to the operation, and the state of each pattern matching position of VS horizontal scanning lines of the horizontal candidate point detection circuit The output of each latch unit corresponding to each of the VS individual inspection coordinate positions of the horizontal candidate point detection circuit is converted into individual inspection coordinate positions so that the data can be ascertained by a single data input by the microcomputer in order to be able to ascertain. If there is at least one latch that is larger than the threshold value in each latch unit corresponding to the A candidate point detection circuit that informs the computer that there is a horizontal correlation line exceeding the correlation threshold value. The first coarse detection is performed at the first pixel interval for both the horizontal image data extraction unit and the vertical image data extraction unit. The first coarse detection is performed by setting the image data to be inspected to be output simultaneously and arranging the central position of the inspection position of VS rows × HS columns to correspond to the center position of the detection range of the image to be inspected. If a candidate point whose output from the candidate point detection circuit is larger than the first coarse detection threshold value set in advance by the microcomputer is found, the horizontal candidate point detection circuit sends the candidate point in advance by the microcomputer. First set
The coordinate position of the highest correlation value of the correlation data of the first plurality of candidate points that has become larger than the coarse detection threshold value for the first time is obtained, and the coordinate position of the highest correlation value obtained in the first coarse detection is obtained. At the center, the second coarse detection is {first pixel interval−second pixel interval × (HS−1) / 2} ≦ 0 in the horizontal direction, and {first pixel interval−second Pixel interval × (VS−
1) / 2｝ ≦ 0 is satisfied at the second pixel interval that satisfies the condition. The candidate point detection circuit makes the microcomputer larger than the second coarse detection threshold value set after the first coarse detection by the microcomputer. The coordinate position of the highest correlation value of the correlation data of the plurality of candidate points in the second time is calculated, and the pattern matching in the third time and thereafter is performed by one pixel with the highest value position coordinate obtained in the second coarse detection as the center. Until performed at intervals, the horizontal direction is
Pixel interval of the first pattern collation−pixel interval of the (P + 1) th pattern collation × (HS−1) / 2｝ ≦ 0, and in the vertical direction, {pixel interval of the Pth pattern collation−P + 1th pattern collation Pixel spacing × (VS-1) / 2｝ ≦ 0 is satisfied at the pixel interval of the (P + 1) th pattern matching that satisfies the condition, and pattern matching is performed at one pixel interval by the microcomputer from the horizontal candidate point detection circuit. The coordinate position having the maximum value of the correlation data of the (P + 1) th single or plural candidate points, which is larger than the detected threshold value for the (P + 1) th time, is set as the detection position. Is done.

According to the present invention, even when the pattern shape of the image in the image to be inspected is monotonous and a similar pattern repeats continuously, the two-dimensional image data to be inspected from the imaging device is output as continuous serial data. Since it is difficult to specify a position by position detection unless a large reference image area of a two-dimensional image is taken, a correlation value at a certain coordinate position may not be calculated unless a large number of pixels are simultaneously processed. Paying attention to one coordinate position, the correlation value at that position is the serial data (=
The serial data (= inspection image data) to be compared with the reference image data) is output simultaneously within one scanning period, and the inspection image data at each corresponding position is synchronized with the video signal scanning period one pixel at a time. By performing the correlation calculation processing, even if it is two-dimensional area data, it is processed as serial data in time, and the correlation calculation of the two-dimensional data is performed by calculating the correlation value at each coordinate position like the processing by one-dimensional correlation calculation. We are simplifying it. That is, instead of completing the correlation value at each coordinate position within one pixel clock period, a large number of portions are calculated in the horizontal and vertical directions to calculate serial data within one scanning period and integrate the result as serial data. At the same time. For this reason, since the size of the reference image does not affect the physical scale of the processing means, it is not necessary to reduce the amount of reference image data during the comparison process by compressing or limiting the original image, and it is not necessary to perform the process. Therefore, even when it is difficult to register a simple and characteristic pattern shape, the position of the pattern can be detected by processing the image data of the original image as it is without accumulating know-how associated with the registration.

Embodiment Hereinafter, an embodiment of a pattern recognition device according to the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a block circuit of the pattern recognition device. As shown in the figure, a pattern recognition device includes a binarization circuit 12 for digitizing a video signal from an imaging device 11, and a binarized inspection image from the binarization circuit 12 for storing a characteristic pattern stored in advance. Reference image storage circuit 16 for pattern matching with reference image
And a variable-length horizontal scanning line shift register capable of holding the binarized image data to be inspected from 1 to 16 lines in the horizontal direction arbitrarily set by an instruction from the microcomputer 13. A vertical image data extraction circuit 14 for simultaneously outputting seven lines of inspected image data evenly spaced at a set interval in the vertical direction of the binarized image with the binarization circuit 12; Variable-length pixels capable of holding the image data to be inspected from 1 pixel to 16 pixels in the horizontal direction by an amount set arbitrarily by an instruction from the microcomputer 13.
By connecting six shift registers in series, the binary image in one horizontal direction is simultaneously output from seven places evenly spaced at a set interval so that the output from the vertical image data extraction circuit 14 is A horizontal image data extraction circuit 15 which connects the binarized image data to be inspected in a row output and simultaneously outputs data at 49 (7 × 7) points of the binarized image, and an image from a reference image storage circuit 16 The pattern matching output of the 49 exclusive OR means for comparing the pattern of the data with the image data to be inspected from the horizontal image data extraction circuit 15 is ANDed with the clock pulse of the pixel frequency. A logic operation circuit 17 that detects the match by generating a pulse corresponding to each pixel and outputs a match pulse of the pixel frequency. It comprises a correlation value accumulating circuit 18 which counts 49 individual pulse outputs, performs pattern matching for each of 49 patterns, and individually accumulates and holds correlation values at each coordinate position.

Next, the operation of each constituent circuit unit will be described in detail. In the actual design example, the field of view of the imaging device 11 is 320 pixels wide,
It is assumed that quantization is performed in a grid of 240 pixels vertically, and that the size of the reference image can be any rectangular range within 128 × 128 pixels. The configuration of the reference image storage circuit 16 is such that a counter for generating operation start and completion timings is provided with a total of four counters in each of the horizontal and vertical directions, and a timing signal for reading or writing the reference image is generated. During the period in which reading or writing of data is valid, a counter that counts a clock corresponding to a pixel is operated to input and output image data in the storage circuit. This makes it possible to arbitrarily change the pattern size of the reference image by changing the set values to the counters that respectively generate the timing signals for starting and completing the operation of the reference image storage circuit 16.

In this embodiment, the size of the reference image is set to a rectangular range of 128 × 128 pixels or less. However, if the storage capacity of the memory is within the range allowed, pattern matching can be performed using a larger reference image. Of course, it can be designed arbitrarily. Moreover, if a memory having a large storage capacity is used, the registrable size per one reference image can be changed without increasing the circuit scale. The increase does not increase significantly.

The memory of the reference image storage circuit 16 is one pixel of 256 kbit etc.
When a RAM serving as an address is used, the horizontal and vertical sizes of the pattern can be set in units of one pixel for the size of the rectangular range for pattern matching. Furthermore, since the storage capacity per pattern is 16 kbits, the pattern setting of the reference image can be used not as a pattern range of 128 × 128 pixels but as a storage capacity of 16 kbits per pattern. For example, a pattern area of 200 x 80 pixels or 100 x 150 pixels, as required, as long as the reference image storage capacity of a rectangle that allows either the horizontal or vertical direction to exceed the length of 128 pixels is allowed It is.

When the memory of the reference image storage circuit 16 is constituted by a × 4 bit or × 8 bit RAM which is not one address per pixel, for example, if a 32k × 8 bit RAM is used, 8 pixels are parallel input / output, and the horizontal pattern setting is 8 bits. It is a pixel unit. It is also possible to set a rectangular reference image so that either the horizontal or vertical direction exceeds the length of 128 pixels if necessary
This is the same as the case where the design is performed using a RAM having one address per pixel such as 256 k × 1 bit.

Further, it is possible to move the generation of the pattern of the reference image to an arbitrary coordinate position by changing the set value in the counter for generating the timing signals for starting and completing the operation of the reference image storage circuit 16, respectively. This means that the coordinate position where the pattern is stored is 2 pixels from the 100th pixel in the horizontal direction.
If it is up to the 00th pixel, the start counter = 100 and the completion counter = 200 may be set and the image data may be written to the memory. Then, the value to be set may be changed and read out while the difference between the start counter value and the completion counter value remains unchanged. That is, start counter = 50, completion counter = 150
If set to, the image data is output at the position shifted by 50 pixels to the left in the horizontal direction. If the start counter is set to 160 and the completion counter is set to 260, the reference image data is output to the position shifted by 60 pixels to the right in the horizontal direction. Then, the reference image data stored at an arbitrary coordinate position is output. Thereby, the input of the inspection image data can be a real-time input,
The pattern matching is performed by matching the output timing of the reference image data with the real-time input. As described above, the function of moving the pattern generation of the reference image to an arbitrary coordinate position is used to achieve the purpose of performing pattern matching in an arbitrary inspection image range of the inspection image.

Next, another embodiment according to the present invention will be described with reference to the drawings. FIG. 2 is a block diagram showing a block circuit of a pattern recognition apparatus according to a second embodiment of the present invention. As shown in FIG. 1, the block diagram includes a binarization circuit 12 for digitizing a video signal from an imaging device 11, and a characteristic pattern of a binarized image from the binarization circuit 12 as a reference image. A reference image storage circuit 16 for storing and pattern matching with the image to be inspected, and the binary image data to be inspected from the binarization circuit 12 from the first row to the 16th row in the horizontal direction are instructed by the microcomputer 13. Six variable-length horizontal scanning line shift registers composed of six variable-length horizontal scanning line shift registers capable of holding an arbitrarily set amount
A vertical image data extraction circuit for outputting seven lines of inspected image data evenly spaced at a set interval in the vertical direction of the binary image by the binarization circuit 20 and the binarization circuit 12 at the same time
14 and the binarized image data to be inspected from the binarization circuit 12 are micro-computered from 1 pixel to 16 pixels in the horizontal direction.
By connecting six variable-length pixel shift registers that can be held arbitrarily set by the instruction from 13 in series, binary images in one horizontal direction are evenly spaced at a set interval at the same time. By providing seven stages to be output from seven places to connect the binarized image data to be inspected of the row output from the vertical image data extraction circuit 14, a total of 7 × 7 (= 49) Horizontal image data extraction circuit 15 that outputs the binarized inspection image data at the point simultaneously
And pattern matching outputs from 49 exclusive OR means for comparing the pattern of the image data from the reference image storage circuit 16 with the image pattern data to be inspected from the horizontal image data extraction circuit 15 and a clock pulse having the same frequency as the pixel. If the pattern matches as a result of the pattern matching, the matching is notified by generating a pulse corresponding to each pixel, and a matching pulse having the same frequency as the pixel is output. An arithmetic circuit 17 and a correlation value integrating circuit 18 that individually integrates and holds the correlation values at each coordinate position where each of the 49 matched patterns has been subjected to pattern matching by individually counting each matched pulse output from the logical arithmetic circuit 17; ,
Further, by comparing and determining the correlation value data of the correlation value accumulating circuit 18 with a detection threshold value set in advance by the microcomputer 13, the outputs of the correlation value accumulating circuit 18 are individually determined so that candidate points can be detected at high speed. The microcomputer 13 is provided with only 49 latches each corresponding to 49 different latches for comparing and judging the magnitude of the detection threshold value and the latch for holding the result. Put together the line comparison results 1
A horizontal candidate point detection circuit 19 is provided, which is provided with only seven latches for vertical pattern matching, which can be grasped by the number of readings.

The operation of each circuit unit thus configured will be described in detail. In the actual design example, the field of view of
It is assumed that quantization is performed into a lattice shape of 320 pixels and 240 pixels vertically, and the size of the reference image can be any rectangular range within 128 × 128 pixels. The configuration of the reference image storage circuit 16 is such that a counter for generating operation start and completion timings is provided with a total of four counters in each of the horizontal and vertical directions, a timing signal for reading or writing the reference image is generated, and the reference timing signal is used for the reference image. By operating a counter that counts a clock corresponding to a pixel during a period in which image reading or writing is valid, image data in the reference image storage circuit is input / output. This makes it possible to arbitrarily change the pattern size of the reference image by changing the set values to the counters that respectively generate the timing signals for starting and completing the operation of the reference image storage circuit 16.

Also in this embodiment, the size of the reference image is a rectangular range of 128 × 128 pixels or less, but can be arbitrarily designed.

Here, the vertical image data extraction circuit 14 outputs the image to be inspected as a video signal from the imaging device 11 and simultaneously outputs the image data to be inspected of a plurality of scanning lines binarized by the binarization circuit 12. Part. This vertical image data extraction circuit 14 will be described with reference to a block circuit diagram shown in FIG. Assuming that the coordinates for pattern matching in one scanning period are 7 points in the horizontal direction and 7 points in the vertical direction, variable-length horizontal scanning lines capable of holding image data to be inspected up to 16 horizontal scanning lines that can be set in units of one horizontal scanning line Six shift registers 20 are required. The six variable-length horizontal scanning line shift registers 20 are connected in series as shown in FIG.
Inspection image data is output from the reference numeral 20. The storage capacity of a single shift register capable of holding image data of an arbitrary length from one horizontal scanning line to 16 horizontal scanning lines for a horizontal scanning line arbitrarily set by an instruction from the microcomputer 13 Is 320 × 16 = 5120 pixels (= 320 pixels in one horizontal scanning line)
5kbit) RAM is required, so generally more than 8kbit RAM
May be used. In order to hold the image data to be inspected for an arbitrary length horizontal scanning line, the lower 9 bits of the RAM address are directly stored in the counter that simply counts the basic clock of the pixel.
And the remaining upper address may be a cyclic counter that clears the count value at an arbitrary set number of times (1 ≦ k ≦ 16) and restarts counting from the beginning. In this way, the seven points V1DT, V2DT, V3DT, V4DT, V5DT, V6DT,
Inspection image data is simultaneously output from V7DT.

Next, the inspection image data from the seven points V1DT, V2DT, V3DT, V4DT, V5DT, V6DT, and V7DT is input to the horizontal image data extraction circuit 15. FIG. 4 is a block diagram showing the configuration of the horizontal image data extraction circuit 15. The pixel data to be inspected having an arbitrary length of up to 16 pixels on the horizontal scanning line is
For the pixel data arbitrarily set by the instruction from (1 ≦ j ≦
16) One variable-length pixel / shift register 21 that can hold only
Six lines are connected in series to the input of the image data to be inspected in the horizontal direction. Since the number of variable-length pixel / shift registers 21 is six, the output is six, but the input image data V1DT to
Since V7DT can also be used as an output, the output from one horizontal line together with this image data is 7 outputs V1DT, V1H
1, V1H2, V1H3, V1H4, V1H5, V1H6. V1DT, V2DT, V3DT, V4DT, V5DT, V6DT, V7 in the vertical direction of the inspected image data
Since each of the six variable-length pixel / shift registers 21 is required independently for each of the DT inputs, a total of 42 variable-length pixel / shift registers 21 are used. As the variable length pixel / shift register 21, SM5828 manufactured by NPC is used. With this IC, up to eight vertical image data extraction circuits
Since image data to be inspected from 14 can be input, 6
By outputting the output from the SM5828 connected in series to the outside, 49 points of image data to be inspected shifted by j pixels in the horizontal direction and by k pixels in the vertical direction are output simultaneously.

The logical operation circuit 17 has 49 reference image data V1DT, V1H1, V1H2, V1H3,... V at each coordinate position from the reference image data in the reference image storage circuit 16 and the horizontal image data extraction circuit 15.
1H6, V2DT, V2H1,..., V7H6 This is a portion for performing an exclusive OR operation on two inputs of one data at a time and outputting the result as a coincidence pulse. The logical operation circuit 17 as shown in the block diagram of FIG. 5 performs exclusive-OR operation simultaneously and independently and independently of 49 types of image data V1DT, V1H1, V1H2, V1H3,. V2H1 ・ ・ ・
A comparison pulse between V7H6 and the reference image data, and the result of the comparison is output as a pulse output to the outside if the two input image data match each other.
P17, P13,... P17, P21, P22,..., P27,..., P37...

The configuration of each logical operation unit 22 will be described with reference to the block diagram of FIG. The logical operation unit 22 is an exclusive OR unit 23 and an exclusive OR unit for performing an exclusive OR operation of two inputs of the image data to be inspected and the reference image data.
A logical AND unit 24 that ANDs the match / mismatch output from 23 with a pixel clock corresponding to one pixel of image data and notifies the outside by outputting a serial signal as a match pulse when the input image data matches. Have been.

In the embodiment of the present invention, the operation of the logical operation unit 22 is configured to output a coincidence pulse which is output to the outside as a pulse output if the two input image data coincide with each other. As a matter of course, it is also possible to perform a pattern matching by searching for a position having a small number of pulses by using a configuration in which an unmatched pulse output to the outside is output as a pulse output.

The correlation value accumulating circuit 18 is constituted by a block diagram shown in FIG. 7, and has 49 independent inspection coordinate positions corresponding to 49 individual inspection coordinate positions which are the result of pattern matching at each coordinate position from the logical operation circuit 17. Matching pulses P11, P12, P13, P17, P21, P22, P27, P37, P7, which are serial signal outputs
It has 49 independent counters 25 for integrating 7 as a correlation value. Since the size of the reference image is within 128 × 128 pixels, in this case, the maximum value of the correlation value is 16384, so that 49 binary counters of 14 bits or more may be used.

Further, the horizontal candidate point detection circuit 19 in the second embodiment compares one correlation value of the correlation value integration circuit 18 with a detection threshold value set in advance by a microcomputer at the same time to determine one horizontal position. The state of the correlation value at seven coordinate positions of the scanning line is determined by one microcomputer 13
This is provided to complete the reading of the state by. FIG. 9 is a block diagram showing the configuration of the horizontal candidate point detection circuit 19. As shown in FIG. Hereinafter, the horizontal candidate point detection circuit 19 will be described with reference to FIG. Horizontal candidate point detection circuit
A counter 19 holds the individual correlation values of the correlation value accumulating circuit 18 and outputs the correlation values as data. A detection threshold holding latch 26 for holding a comparison value by writing data, and a correlation value accumulating circuit 18
Output D11, D12, D13 ... D17, D21, D22 ...
The D27,..., D37,..., D77 independently compare the magnitude of the detection threshold value held in the detection threshold holding latch 26 with the detection threshold value, and hold the result.
49 detection circuits whose status can be grasped at 13.
Consists of 27. When the 49 detection circuits 27 are put together for each of the seven horizontal comparison lines, one comparison line holds the state of the result of pattern matching at seven points. Since each of the results at these seven points is a result of comparing the magnitude with the detection threshold value, each result is represented by 1 bit.
And the amount of information can be 7 bits per horizontal comparison line. By assigning this 7-bit information as one I / O address, one microcomputer 13
By reading I / O, it is possible to grasp the state of pattern matching at seven points. The configuration of each detection circuit 27 will be described with reference to the block diagram of FIG. The detection circuit 27 includes a digital comparator 28 for comparing two-input data of a detection threshold value and a correlation value, and a digital comparator 28
And a result holding latch 29 for holding the result of the magnitude comparison output from. When the horizontal candidate point detection circuit 19 is not built-in, the correlation value data is read seven times to grasp the state of the correlation value at the seven coordinate positions, and the magnitude is compared with the coarse detection threshold value by the microcomputer 13 seven times. However, by incorporating the horizontal candidate point detection circuit 19, one reading of I / O data is equivalent to performing the above processing contents, so that the processing time can be shortened accordingly. I can do it. In order to provide the means for detecting horizontal candidate points in each of the seven horizontal scanning lines, the state of the horizontal candidate point detection circuit 19 is read seven times and the correlation value data is read 49 times, and the microcomputer 13 performs 49 times. The size comparison with the coarse detection threshold value will be performed on behalf of
The specification of the pattern collation position by comparing the magnitude with the detection threshold value can be performed at a higher speed. As described above, it is possible to read the state by the microcomputer 13 at one time to read the state of the correlation value at seven coordinate positions of one horizontal scanning line. Is that the state of the correlation value at 7 coordinate positions of two horizontal scanning lines can be grasped by one data reading.
It is possible to grasp the state of the correlation value at the seven coordinate positions of the horizontal scanning line. In this way, the number of grasps of the state of the correlation value that can be obtained by one data reading is, of course, a plurality of horizontal scanning lines by the microcomputer 13 used for control and the design method of the connection configuration thereof. When the number of pattern matching points in one horizontal scanning line is increased, when a 16-bit microcomputer 13 is used, the state of the correlation value up to 16 coordinate positions can be grasped by a single data read by using a 32-bit microcomputer. When the computer 13 is used, the state of the correlation value up to 32 coordinate positions can be grasped by one data reading. And 16bit and 32bit
When the microcomputer 13 is used, the number of pattern matchings in the horizontal and vertical directions from the total number of pattern matching points is set to various conditions such as the recognition accuracy and inspection range of the recognition target for the purpose of designing in consideration of the above points. It is also possible to perform it together.

In the apparatus according to the present embodiment, the total number of horizontal / vertical pattern matching is set to 7 × 7 = 49 points.
It is of course possible to set an arbitrary design value such as × 5,9 × 9,11 × 7,5 × 15, etc. Obviously, more pattern matching points should be provided to complete the pattern matching process. Increasing the number of pattern matching points is advantageous because the pixel interval for matching can be narrowed and the probability of completing pattern matching with a smaller number of pattern matches is advantageous.

The recognition apparatus of the present invention is one of the coarse and fine detection methods, and a method of determining a pixel interval when performing coarse detection will be described below as to how to shift from coarse detection to precise detection. In the first coarse detection, the horizontal image data extraction circuit 15 and the vertical image data extraction circuit 14 are both inspected image data V1DT, V1H1, V1H2, V1H3,... V1H6, V2DT, V
·································································································································································· First
The second coarse detection is performed, and the second detection is performed using the coordinate position corresponding to the portion holding the maximum correlation value in the correlation count value based on the correlation values of the position data of the first 49 candidate points. Is a coordinate position where When the next detected pixel interval is selected so that the next inspection image range includes the pixel interval of the coarse detection performed first in both the horizontal and vertical directions, the pattern matching for the second and subsequent pattern matchings is performed. It is assumed that the coordinate position where the correlation value becomes the maximum is included in the inspection image range and is not overlooked on the assumption that the inspection range is gradually narrowed.

Therefore, in general, the pixel interval of the p-th pattern matching in the horizontal direction is dkp, the pixel interval of the p + 1-th pattern matching is dk (p + 1), and the number of points of the pattern matching in one horizontal scanning line is an odd value HS. For the pixel interval dk (p + 1) of the (p + 1) -th pattern matching, select the smallest value among values satisfying dkp−dk (p + 1) × (HS−1) / 2 ≦ 0 (1) Is an efficient selection method for converging the inspection image range with a smaller number of coarse detection processes. The same applies to the method of efficiently selecting the pixel interval for pattern matching in the vertical direction.

Therefore, in the case of the present embodiment, since dk1 = 8 and HS = 7, the matching pixel interval for performing the second pattern matching is obtained.
From equation (1), dk2 is calculated as follows: 8-dk2 × (7-1) / 2 ≦ 0 (2) (3) Therefore, from the minimum integer value satisfying the above expression (3), dk
2 = 3 is obtained. Since the matching pixel interval dk3 for performing the third pattern matching is dk2 = 3 and HS = 7, the expression (1)
And dk3 = 1 is obtained in the same manner. The second pattern matching pixel interval dj2 in the vertical direction is dj1 = 8, VS =
7 is the same as that in the horizontal direction, dj2 = 3, and the matching pixel interval dj3 for performing the third pattern matching is dj2 =
3. Since VS = 7, dj3 = 1 is similarly obtained from equation (1). The detection operation will be described below.

Horizontal image data extraction circuit 15 for the first coarse detection
Microcomputer to the vertical and vertical image data extraction circuit 14
From 13, data is set so that the inspection image data is output at an interval of 8 pixels in both the horizontal and vertical directions. Also, the output position of the reference image data is set to the microcomputer so that the center position (4th row, 4th column) of each inspection position for pattern matching of 7 rows and 7 columns is arranged at the center position of the inspection range of the image to be inspected.
Controlled by 13. Controlling the output position of the reference image data so that the center position (4th row and 4th column) of each inspection position of pattern matching is arranged is, specifically, the variable length of 4th row and 4th column shown in FIG. In the inspection image data output from the pixel / shift register 21, the partial image 6 shown in FIG.
Is located at the center position to be inspected, and the first image of the reference image 7 is synchronized with the scanning timing at which the image data of the first row and first column of the arranged partial image 6 of m rows and n columns is output.
That is, the image data in the first column of the row is output.
Therefore, the same m rows n
This means that the partial images 6 in the column are arranged at intervals of 8 pixels.

Next, all the count values of the 49 correlation value accumulating circuits 18 are cleared and pattern matching is started. After the first pattern matching, the correlation data of each counter 25 in the correlation value accumulating circuit 18 is read, the correlation data having the maximum value is searched, and the pattern matching coordinate position where the maximum correlation value is held. Is calculated, and if the value is larger than the coarse detection threshold value at an interval of 8 pixels, the pattern matching coordinate position holding the maximum correlation value is set as the coordinate position for performing the second detection. Next, as preparation for the second pattern matching, the microcomputer 13 outputs image data to be inspected to the horizontal image data extraction circuit 15 and the vertical image data extraction circuit 14 at three pixel intervals in both the horizontal and vertical directions. And the output position of the reference image data is set to the center position of each inspection position (4th row, 4th column) so that pattern matching is performed centering on the coordinate position calculated in the first pattern matching processing. Are set so as to be arranged, the detection threshold value is set, the count values of the 49 correlation value integrating circuits 18 are all cleared, and the second pattern matching is started.

After the second pattern matching, the correlation data is read, the data having the maximum value is searched for, and the pattern matching coordinate position holding the maximum correlation value is calculated. Next, as preparation for the third pattern matching, the microcomputer 13 outputs image data to be inspected to the horizontal image data extraction circuit 15 and the vertical image data extraction circuit 14 at one pixel intervals in both the horizontal and vertical directions. The output position of the reference image data is set to the center position of each inspection position (4th row, 4th column) so that pattern matching is performed centering on the coordinate position calculated in the second pattern matching processing. Are set, the detection threshold value is set, the count values of the 49 correlation value integrating circuits 18 are all cleared, and the third pattern matching is started.

After the third pattern matching, the data in the correlation value accumulating circuit 18 is read by the microcomputer 13 to search for the data having the maximum value among the data, and the maximum correlation value is larger than a predetermined detection threshold value. If so, the pattern matching coordinate position corresponding to the portion holding the maximum correlation value is calculated, and the coordinate position is set as the detection position.

In the second embodiment, 49 correlation value integrating circuits 18
Clear all the count values of the horizontal candidate point detection circuit 19
The threshold value for coarse detection at an interval of 8 pixels is set in the detection threshold holding latch 26 of FIG.
Is cleared and pattern matching starts.

After the first pattern matching, the horizontal candidate point detection circuit 19
The contents of the seven result holding latches 29 are read and the counters 25 in the correlation value accumulating circuit 18 which hold the correlation values larger than the coarse detection threshold value at the interval of 8 pixels are specified. The correlation data of the counter 25 is read, the correlation data having the maximum value is searched for, the coordinate position corresponding to the pattern collation coordinate position holding the maximum correlation value is calculated, and the coordinates for performing the second detection are obtained. Position. Next, in preparation for the second pattern matching,
Horizontal image data extraction circuit 15 and vertical image data extraction circuit
The microcomputer 13 sets data so that the image data to be inspected is output at three pixel intervals in both the horizontal and vertical directions in accordance with the above-described pixel interval determination method, and the coordinates calculated in the first pattern matching process. The output position of the reference image data is controlled so that the center position (4th row, 4th column) of each inspection position is arranged so that pattern matching of 7 rows and 7 columns is performed centering on the position, and 49 correlations are obtained. All the count values of the value accumulating circuit 18 are cleared, the threshold value for coarse detection at three pixel intervals is set in the detection threshold holding latch 26 of the horizontal candidate point detection circuit 19, and the result holding latch 29 for all 49 is cleared. , The second pattern matching is started.

After the second pattern matching, the horizontal candidate point detection circuit 19
The contents of the seven result holding latches 29 are read and the counters 25 in the correlation value accumulating circuit 18 which hold the correlation values larger than the coarse detection threshold value at three pixel intervals are specified. The correlation data of the counter 25 is read, the data having the maximum value is searched for, the coordinate position corresponding to the pattern collation coordinate position holding the maximum correlation value is calculated, and the coordinate position for performing the third detection is obtained. And Next, as preparation for the third pattern matching, the microcomputer 13 outputs the image data to be inspected to the horizontal image data extraction circuit 15 and the vertical image data extraction circuit 14 at one pixel intervals in both the horizontal and vertical directions. And the output position of the reference image data is set at the center position (4) of each inspection position so that pattern matching of 7 rows and 7 columns is performed centering on the coordinate position calculated in the second pattern matching process. (4th row and 4th column) are arranged, and the detection threshold value at three pixel intervals is set in the detection threshold holding latch 26 of the horizontal candidate point detection circuit 19, and 49 correlation value integration circuits are set.
Clear all 18 count values and start the third pattern matching.

After the third pattern matching, the horizontal candidate point detection circuit 19
The contents of the seven result holding latches 29 are read and the counters 25 in the correlation value accumulating circuit 18 which hold the correlation values larger than the coarse detection threshold value at three pixel intervals are specified. The correlation data of the counter 25 is read, the data having the maximum value is searched for, the coordinate position corresponding to the pattern matching coordinate position holding the maximum correlation value is calculated, and the coordinate position is set as the detection position.

Next, a third embodiment of the present invention will be described with reference to FIG.

In the first embodiment, the data at the horizontal candidate point detection circuit 19 is read by the microcomputer 13 to ascertain the state at the pattern matching position, and the maximum value of the correlation value is searched. FIG. 8 shows a block diagram of the recognition device in this embodiment. The second embodiment differs from the first embodiment shown in FIG. 1 only in that a candidate point detection circuit 10 is additionally provided, and the other components are the same. Therefore, the binarization circuit 12, the vertical image data extraction circuit 14, the horizontal image data extraction circuit 15, the reference image pattern storage circuit 16, the logical operation circuit 1
7. The description of the correlation value accumulating circuit 18 is omitted because it is the same as that of the first embodiment.

The candidate point detection circuit 10 is provided for each of the seven horizontal comparison lines of the horizontal candidate point detection circuit 19 so that the state of the correlation value data at a plurality of positions of the horizontal candidate point detection circuit 19 can be grasped by one data input. If at least one of the seven result comparison latches 29 has a value larger than the value set in the detection threshold value holding latch 26 in the output of the result holding latch 29, the state of each of the result comparison latches 29 is determined. Microcomputer by holding as information in comparison line
The pattern matching coordinate position exceeding the detection threshold
This is provided in order to grasp the state of the entire coordinate position where pattern matching has been performed by the microcomputer 13 with less data having been in 49 points as a more aggregated information. FIG. 11 is a block diagram showing the configuration of the candidate point detection circuit 10. As shown in FIG. The outputs Q11 and Q12 of the result holding latch 29 from each of the seven horizontal comparison lines
Q13, Q17, Q21, Q22, Q27, Q37, Q77 hold a correlation value exceeding the detection threshold value in each horizontal comparison line by ORing 7 bits in each horizontal comparison line as a unit. The presence / absence of the presence / absence is determined by using the information on one horizontal comparison line as 1-bit information and holding the states of the seven horizontal comparison lines in the seven line detection circuits 30. That is, as shown in FIG. 11, (Q11, Q12, Q13..Q
17), (Q21, Q22, Q23..Q27), (Q31, Q32, Q33..Q3
7) The 7-bit inputs of (Q71, Q72, Q73..Q77) are logically ORed respectively, and the result is further divided by seven line detection circuits 30 into 1
The information of the horizontal comparison line is held as 1-bit information.
In the present embodiment, since seven horizontal comparison lines are provided, the information of each horizontal comparison line can be 7 bits, and detection of all pattern collation positions by one reading of the I / O data 7 by the microcomputer 13 is performed. It is possible to grasp the presence or absence of a coordinate position exceeding the threshold value.

The candidate point detection circuit 10 further summarizes the state of the correlation value at the 49-point coordinate position from the horizontal candidate point detection circuit 19, and displays the state of the correlation values of all 49 inspection ranges in the pattern matching. This is provided to perform high-speed comparison and judgment processing which can be grasped by one data read by the microcomputer 13 and performed by software by the microcomputer 13. As stated in
When the 16-bit microcomputer 13 is used, the state of the correlation value at seven coordinate positions of two horizontal scanning lines can be grasped by one data reading, and when the 32-bit microcomputer 13 is used, four times can be obtained by one data reading. It is possible to grasp the state of the correlation value at the seven coordinate positions of the horizontal scanning line. The microcomputer used for control in this way
According to the design of 13 and its connection configuration, the number of correlation value states that can be obtained by one data reading is of course a plurality of horizontal scanning lines.

In the apparatus according to the present embodiment, the total number of horizontal / vertical pattern matching is set to 7 × 7 = 49 points, but 9 × 9,11 ×
Of course, it is possible to set an arbitrary design value such as 7,5 × 15, etc. It is clear that the pattern matching point should be provided. Increasing the number of pattern matching points is advantageous because the probability of completing pattern matching with a smaller number of pattern matchings is increased.

In the present invention, how to shift from coarse detection to more precise detection, and the method of determining the pixel interval when performing coarse detection is the same as in the first embodiment, so that the description of this case is omitted. .

Hereinafter, the detection operation will be described. For the first coarse detection, the microcomputer 13 sets data to the horizontal image data extraction circuit 15 and the vertical image data extraction circuit 14 so that the image data to be inspected is output at eight pixel intervals in both the horizontal and vertical directions. . Also, the output position of the reference image data is set to the microcomputer 13 so that the center position (4th row, 4th column) of each inspection position for pattern matching of 7 rows and 7 columns is arranged at the center position of the inspection range of the image 5 to be inspected. Is controlled by Controlling the output position of the reference image data so that the center position (4th row and 4th column) of each inspection position for pattern matching is arranged specifically means the variable length of 4th row and 4th column shown in FIG. In the image data to be inspected output from the pixel shift register 21, it is assumed that the partial image 6 in FIG. 2 is arranged at the center position to be inspected, and the first row and first column of the arranged 128-row × 128-column partial image 6 That is, the image data of the first row and the first column of the reference image data 7 is output in synchronization with the scanning timing at which the image data of the eye is output. Accordingly, the same partial image 6 of 128 rows × 128 columns is disposed at an interval of 8 pixels on the upper, lower, left and right sides of the disposed partial image 6. Next, the count values of the 49 correlation value accumulating circuits 18 are all cleared, and the detection threshold holding latch 26 of the horizontal candidate point detection circuit 19 is set with the coarse detection threshold value at 8-pixel intervals. The holding latch 29 is cleared, the candidate point detection circuit 10 and all the 7-bit line detection circuits 30 are cleared, and pattern matching is started.

After the first pattern matching, the microcomputer 13 reads the contents of the line detection circuit 30 in the candidate point detection circuit 10 and there is a candidate point at a detection position exceeding the first coarse detection threshold value set in advance. The information of the corresponding result holding latch 29 in the horizontal candidate point detection circuit 19 of the horizontal comparison line indicating the existence of the horizontal comparison line is read if it exists, and the correlation at the pattern matching coordinate position corresponding to the information is read. From the counter 25 holding the value, the largest correlation value counter 25 holding the correlation value larger than the coarse detection threshold value among the correlation values of the plurality of corrections performed in the first pattern matching. The pattern matching coordinate position corresponding to the part holding the correlation value is set as the center position of the next pattern matching. Next, as preparation for the second pattern matching, the microcomputer 13 outputs image data to be inspected to the horizontal image data extraction circuit 15 and the vertical image data extraction circuit 14 at three pixel intervals in both the horizontal and vertical directions. The data is set, and the output position of the reference image data is set to the center position (4th row, 4th column) of each inspection position so that pattern matching is performed centering on the coordinate position calculated in the first pattern matching processing. Control so that they are arranged, set the detection threshold value, clear all count values of the 49 correlation value accumulating circuits 18, and coarsely detect the detection threshold holding latch 26 of the horizontal candidate point detection circuit 19 at three pixel intervals. A threshold value is set, all 49 result holding latches 29 are cleared, all 7-bit line detection circuits 30 of the candidate point detection circuit 10 are cleared, and the second pattern matching is started.

After the second pattern matching, the microcomputer 13 reads the contents of the candidate point detection circuit 10 and checks whether there is a candidate point at a detection position exceeding the first coarse detection threshold value set in advance. The information of the corresponding result holding latch 29 in the horizontal candidate point detection circuit 19 of the horizontal comparison line indicating the presence is read in the case where it is present, and the correlation value at the pattern matching coordinate position corresponding to the information is held. From the counter 25, the maximum correlation value is held in the correlation value counter 25 holding the correlation value larger than the coarse detection threshold value among the correlation values of the plurality of candidate points subjected to the second pattern matching. The pattern matching coordinate position corresponding to the portion is set as the center position of the next pattern matching. Next, as preparation for the third pattern matching, the microcomputer 13 outputs image data to be inspected to the horizontal image data extraction circuit 15 and the vertical image data extraction circuit 14 at three pixel intervals in both the horizontal and vertical directions. Data is set, and the output position of the reference image data is set to the center position (4th row, 4th column) of each inspection position so that pattern matching is performed around the coordinate position calculated in the first pattern matching processing. Control so that they are arranged, set the detection threshold value, clear all count values of the 49 correlation value accumulating circuits 18, and coarsely detect the detection threshold holding latch 26 of the horizontal candidate point detection circuit 19 at one pixel intervals. Set the threshold value for
Candidate point detection circuit by clearing all 49 result holding latches 29
The 10th 7-bit line detection circuit 30 is all cleared, and the third pattern matching is started.

After the third pattern matching, the microcomputer 13 reads the contents of the candidate point detection circuit 10 and checks whether there is a candidate point at a detection position exceeding the first coarse detection threshold value set in advance. The information of the corresponding result holding latch 29 in the horizontal candidate point detection circuit 19 of the horizontal comparison line indicating the presence is read in the case where it is present, and the correlation value at the pattern matching coordinate position corresponding to the information is held. The counter 25 holds the maximum correlation value in the correlation value counter 25 that holds a correlation value larger than the coarse detection threshold value among the correlation values of a plurality of candidate points subjected to the third pattern matching. A coordinate position corresponding to the pattern matching coordinate position is calculated, and the coordinate position is set as a detection position.

Effect of the Invention As described above, according to the pattern recognition method or apparatus of the present invention, the size of the reference image can be arbitrarily set, and the position of the recognition target, which has been almost impossible in the past, can be detected. Furthermore, since it is possible to set a reference image pattern of an arbitrary size, the work of registration is facilitated, and even if pattern matching is performed on a large image area to be inspected, position detection can be performed quickly without increasing processing time. Became possible.

[Brief description of the drawings]

FIG. 1 is a block diagram showing the configuration of a first embodiment of a pattern recognition apparatus according to the present invention, FIG. 2 is a block diagram showing the configuration of the second embodiment, and FIG. FIG. 4 is a block diagram illustrating a configuration of a vertical image data extraction circuit according to the embodiment; FIG. 4 is a block diagram illustrating a configuration of a horizontal image data extraction circuit; FIG. FIG. 6 is a block diagram illustrating the configuration of each logical operation unit of the logical operation circuit,
FIG. 7 is a block diagram illustrating the configuration of the correlation value integrating circuit, FIG. 8 is a block diagram illustrating the configuration of the third embodiment of the present invention, and FIG. 9 is a horizontal candidate in the embodiment of FIG. Block diagram illustrating the configuration of the point detection circuit, FIG. 10 is a block diagram similarly illustrating the configuration of each detection circuit of the horizontal candidate point detection circuit, FIG. 11 is a block diagram similarly illustrating the configuration of the candidate point detection circuit, FIG. 12 is a schematic diagram of a pattern matching process which is a premise of the present invention. 10: Candidate point detection circuit, 11: Image pickup device, 12: Binarization circuit, 13: Microcomputer, 14: Vertical image data extraction circuit, 15: Horizontal image data extraction circuit, 16: Reference Image storage circuit, 17 ... Logic operation circuit, 18 ... Correlation value integration circuit, 19 ... Horizontal candidate point detection circuit.

Claims (4)

(57) [Claims] 1. Reference image storage means for digitizing a video signal of an image pickup means for picking up an image to be inspected, extracting a characteristic pattern at an arbitrary size, and outputting image data previously stored in a memory as a reference image; The number (VS-1) of horizontal scanning line holding means capable of holding an arbitrary number dy of horizontal scanning lines of the two-dimensional inspection image data by the number of horizontal scanning lines set by the microcomputer is set to a horizontal scanning line interval k (1 ≦ 1). k ≦ dy) vertical image data extracting means for simultaneously outputting inspected image data corresponding to the inspection coordinates of a plurality of rows VS evenly spaced, and an inspected image from the vertical image data extracting means A series of (HS-1) pixel shift means capable of holding the simultaneous output of data up to an arbitrary number of pixels dx in the horizontal direction by the number of pixels set by the microcomputer. It is to connect, the pixel j of inspection image data of one horizontal direction is a certain set time
(1.ltoreq.j.ltoreq.dx) The ones that are simultaneously output from the HS locations evenly spaced at intervals are output from the vertical extraction means.
By inputting the image data to be inspected simultaneously output from the VS horizontal scanning lines as inputs, and providing the (HS-1) serial pixel shift means in the VS stage, the total of the two-dimensional image data to be inspected can be obtained.
Horizontal image data extracting means for simultaneously outputting the inspected image data corresponding to the individual inspection coordinate positions of the VS × HS points; and from the reference image data and the horizontal image data extracting means from the reference image storage means. Inspection image data corresponding to each inspection coordinate position is compared at VS x HS points at the same time, and the pattern matching output corresponding to each inspection coordinate position at VS x HS points matches VS x HS at the same time A logical operation means for simultaneously outputting serial data from each part corresponding to each inspection coordinate position as data, and simultaneously counting each matched data output corresponding to each inspection coordinate position from the logical operation means. By
VS × HS pattern matching is completed by one image scan, and is constituted by correlation value accumulating means for independently calculating and holding as a correlation value in a portion corresponding to each inspection coordinate position. The first coarse detection is performed on horizontal image data. The extraction means and the vertical image data extraction means are set so that HS × VS inspection image data are simultaneously output at the pixel interval of the first pattern matching, and the HS row × HS × which is arranged so that the center of the inspection position of the VS row corresponds to the first coarse detection and the first pattern matching is performed
Search for the part holding the maximum correlation value in the correlation data of the VS candidate points, and if the value is larger than a predetermined coarse detection threshold value, the pattern matching coordinate position corresponding to the part holding the maximum correlation value Is determined as the center coordinate position of the next pattern matching, and the center of the inspection position of HS rows × VS columns is centered on the coordinate position of the highest correlation value obtained in the first coarse detection. In the horizontal direction, the second coarse detection is performed by dividing the first pixel interval−the second pixel interval × (HS−
1) / 2} ≦ 0, and in the vertical direction, a second pixel interval that satisfies {first pixel interval−second pixel interval × (VS−1) / 2} ≦ 0 and is a minimum integer value. 2nd HS × VS
The coordinate position of the highest correlation value of the correlation data of the candidate points is obtained, and the center of the HS row × VS column is arranged around the coordinate position of the highest correlation value obtained in the second coarse detection. For the third and subsequent pattern comparisons, the set pixel interval in the horizontal and vertical directions is 1
Until the pixel interval is reached, the horizontal direction is {P-th pattern matching pixel interval−P + 1-th pattern matching pixel interval ×
(HS-1) / 2} ≦ 0, and the vertical direction satisfies {P-th pattern matching pixel interval−P + 1-th pattern matching pixel interval × (VS−1) / 2} ≦ 0. At the pixel interval of the (P + 1) -th pattern matching, which is an integer, the center of the HS row × VS column is arranged at the coordinate position of the P-th highest correlation value, and the pattern matching performed at the 1-pixel interval is performed. If the maximum value of the correlation data of the HS × VS candidate points is larger than a predetermined detection threshold value, the pattern recognition method uses the coordinate position corresponding to the portion holding the maximum correlation value as the detection position. 2. A reference image storage means for digitizing a two-dimensional video signal from an imaging means, extracting an arbitrary size of a characteristic pattern of an image to be inspected, and outputting image data previously stored in a memory as a reference image, A (VS-1) number of horizontal scanning line holding means capable of holding an arbitrary number dy of horizontal scanning lines of the two-dimensional inspection image data by an amount set by the microcomputer has a horizontal scanning line interval k (1 ≦ k ≦ dy). Vertical image data extracting means for simultaneously outputting the inspected image data corresponding to the individual inspected image positions of the VS rows evenly spaced by the above); and the inspected image data from the vertical image data extracting means. of
By connecting in series (HS-1) pixel shift means capable of holding VS simultaneous outputs up to an arbitrary pixel dx in the horizontal direction by the number of pixels set by the microcomputer, an image to be inspected in one horizontal direction is obtained. Data obtained by simultaneously outputting data from HS locations evenly spaced at a set pixel j (1 ≦ j ≦ dx) interval from the vertical image data extraction means from VS horizontal scanning lines With the image data to be inspected output at the same time as input, by providing the (HS-1) serial pixel shift means in the VS stage, it is possible to correspond to each inspection coordinate position of the statistical VS × HS point of the two-dimensional image data to be inspected. Horizontal image data extracting means for simultaneously outputting image data to be inspected, image data from the reference image storing means, and inspected parts corresponding to individual inspection coordinate positions from the horizontal image data extracting means. VS × for pattern comparison image pattern data at the same time
A logical operation consisting of outputting pattern matching output corresponding to each inspection coordinate position composed of HS exclusive ORs as VS × HS simultaneous coincidence pulses from the exclusive OR means corresponding to each inspection coordinate position. VS × HS pattern matching is completed in one image scan by simultaneously counting each coincident pulse output corresponding to each inspection coordinate position from the logical operation means and corresponding to each inspection coordinate position. A correlation value integrating means for independently calculating and holding the calculated value, and a pattern at a candidate point by simultaneously comparing and determining the correlation value data of the correlation value integrating means with a pattern detection threshold value set in advance by a microcomputer. The data output of the correlation value accumulating means is compared with the held detection threshold value so that a matching state can be detected. At the same time, the magnitude is compared and determined between the corresponding coordinate positions, and the holding means corresponding to each inspection coordinate position for holding the result is provided only for the number corresponding to VS × HS individual inspection coordinate positions, The microcomputer uses the means for compiling the comparison results of one horizontal scanning line HS at a time to maintain the state so that the state of each pattern matching position of one horizontal scanning line can be grasped by one reading of the microcomputer. By providing only VS lines, it is constituted by horizontal candidate point detecting means capable of grasping the state of each pattern matching position of each horizontal scanning line and specifying the position corresponding to the calculation, and the first coarse detection The image data extraction means / vertical image data extraction means simultaneously output VS × HS inspection image data at the pixel interval of the first pattern matching. And sea urchin set, the center position of the detection range of the inspection image
The first coarse detection is performed by arranging the inspection positions of VS rows × HS columns so as to correspond to each other, and is set in advance by the microcomputer after the first coarse detection by the horizontal candidate point detecting means. It checks whether there is a candidate point at the detection position exceeding the first coarse detection threshold value, and reads the information of the horizontal candidate point detection means of the horizontal comparison line indicating the existence if it exists, The means for holding the correlation value at the pattern matching coordinate position corresponding to the information corresponds to the part holding the maximum correlation value among the correlation count values of the plurality of candidate points for which the first pattern matching was performed. The pattern matching coordinate position is determined as the center coordinate position of the next pattern matching coordinate, and the center of the inspection position of VS rows × HS columns centering on the coordinate position of the highest correlation value obtained in the previous first coarse detection. Corresponds Place manner, the horizontal direction of the second grain roughness detection {first pixel interval - the second pixel interval ×
(HS-1) / 2} ≦ 0, and the vertical direction is {first pixel interval−
The second pixel interval × (VS−1) / 2｝ ≦ 0 is arranged such that the center of the VS row × HS column corresponds to the second pixel interval satisfying the second pixel interval × (VS−1) / 2｝ ≦ 0. The computer obtains the coordinate position of the highest correlation value of the correlation data of a plurality of second candidate points that has become larger than the second coarse detection threshold value set after the first coarse detection and obtains the second coordinate value. Centering on the highest value position coordinates obtained in the second coarse detection
The horizontal direction is the pixel interval of the P-th pattern matching minus the pixel interval of the P + 1-th pattern matching until the center part of the VS row × HS column is arranged and the third and subsequent pattern matchings are performed at one-pixel intervals. × (HS-1) / 2} ≦ 0, the vertical direction is {Pth
The pixel interval of the first pattern matching−the pixel interval of the (P + 1) th pattern matching × (VS−1) / 2｝ ≦ 0.
The center of VS rows × HS columns is arranged at the pixel interval of the + 1st pattern matching, and the microcomputer performs pattern matching at the 1st pixel interval from the horizontal candidate point detecting means by the microcomputer. A pattern recognition method characterized by using, as a detection position, a coordinate position having a maximum value of correlation data of a (P + 1) th single or a plurality of candidate points that has become larger than a threshold value. 3. A binarizing circuit for digitizing a two-dimensional video signal of an image pickup apparatus for picking up an image to be inspected, and a characteristic pattern cut out in an arbitrary size within a memory storage capacity range and previously stored in a memory as a reference image. A reference image storage circuit for outputting image data; and a horizontal scan capable of holding an arbitrary number of horizontal scan lines of the binarized image data to be inspected from the binarization circuit, arbitrarily set by an instruction from the microcomputer. The image data to be inspected corresponding to the inspection coordinate positions evenly spaced at an arbitrary horizontal scanning line interval set in the vertical direction of the binarized image by the binarization circuit, which is constituted by a line shift register, simultaneously with the binarization circuit. A vertical image data extraction circuit for outputting, and a microcomputer for outputting an arbitrary pixel in the horizontal direction from the output of the binary image data to be inspected from the vertical image data extraction circuit. By connecting a plurality of variable-length pixels and shift registers that can hold only the pixels set arbitrarily according to these instructions in series, one horizontal binarized image data to be inspected is simultaneously distributed at a set pixel interval. The serial variable-length pixel shift register is provided with the binary inspected image data simultaneously output from the horizontal scanning line from the vertical extraction circuit as an input so as to be simultaneously output from a plurality of locations spaced apart from each other. A horizontal image data extraction circuit configured to simultaneously output binarized image data to be inspected corresponding to an inspection coordinate position at each point of the binarized image; reference image data from the reference image storage circuit and the horizontal image Each of the inspection coordinate positions formed by an exclusive OR circuit for simultaneously performing pattern comparison of the inspected image pattern data corresponding to the individual inspection coordinate positions from the data extraction circuit. A logical operation circuit for simultaneously outputting the matched pattern matching output as a coincidence pulse from the exclusive OR unit corresponding to each inspection coordinate position, and each coincidence pulse corresponding to each inspection coordinate position from the logical operation circuit A correlation value integration circuit that completes a plurality of pattern comparisons in one field by simultaneously counting the outputs individually, and independently calculates and holds a portion corresponding to each inspection coordinate position, includes a specific pattern in a two-dimensional image. A pattern recognition device characterized by automatically extracting a coordinate position and using a coordinate position as a detection position. 4. A binarizing circuit for digitizing a two-dimensional video signal from an imaging device, and a characteristic pattern of an image to be inspected is arbitrarily sized within a storage capacity range of the memory and cut out and stored in a memory in advance as a reference image. A reference image storage circuit capable of outputting the converted image data, and holding the binarized image data to be inspected from the binarization circuit up to an arbitrary horizontal scanning line by an amount arbitrarily set according to an instruction from the microcomputer. (VS-1), which is composed of possible horizontal scanning line shift registers (VS-1).
1) The horizontal scanning line shift register (1 ≦ k ≦ dx) interval (where k is a pixel interval) set in the vertical direction of the binarized image by the horizontal scanning line shift registers and the second binarization circuit are equal. A vertical image data extraction circuit configured to simultaneously output inspected image data corresponding to individual inspection coordinate positions of a plurality of rows VS, and a binarized image data from the vertical image data extraction circuit. (HS-1) serially connected variable-length image shift registers capable of holding VS simultaneous output of inspection image data up to an arbitrary pixel in the horizontal direction by an arbitrary number of pixels set by an instruction from a microcomputer As a result, the image data to be inspected which has been binarized in one horizontal direction is simultaneously present in a set pixel (1
.Ltoreq.j.ltoreq.dy, which are simultaneously output from the HS locations evenly spaced at intervals (where j is the pixel interval) are simultaneously output from the VS horizontal scanning lines from the vertical image data extraction circuit. The (HS-1) serial variable-length image shift registers are connected to the VS
A horizontal image data extraction circuit for simultaneously outputting binarized image data to be inspected corresponding to individual inspection coordinate positions of a total VS × HS point of the binary image by providing a stage; and storing the reference image. The image data from the circuit and the image data to be inspected corresponding to the individual inspection coordinate positions from the horizontal image data extraction circuit are simultaneously subjected to pattern comparison. A logical operation circuit that outputs the VS × HS matched pattern outputs as simultaneous coincidence pulses from an exclusive OR unit corresponding to each inspection coordinate position, and corresponds to each inspection coordinate position from the logical operation circuit. By simultaneously counting each matched pulse output individually,
A correlation value integrating circuit that completes VS × HS pattern matching in one field and independently calculates and holds the portion corresponding to each inspection coordinate position, and sets in advance the correlation value data of the correlation value integrating circuit from a microcomputer. The output of the correlation value integration circuit is provided with a comparator corresponding to each inspection coordinate position so that a pattern matching state at a candidate point can be detected by comparing and judging the magnitude of the detected pattern detection threshold value and holding the detection threshold value. At the same time, the magnitudes are compared and determined with the latches corresponding to the individual inspection coordinate positions, and the latches corresponding to the individual inspection coordinate positions that hold the results are provided in an amount corresponding to the VS × HS inspection coordinate positions. The microcomputer collects the comparison results of one horizontal scanning line HS location and collects one horizontal scanning line by one reading of the microcomputer. By providing only VS latches for vertical pattern matching, which are capable of ascertaining the pattern matching state of each scanning line, it is possible to ascertain the pattern matching state of each horizontal scanning line and to specify the position corresponding to the operation. A horizontal candidate point detection circuit, and a single data input by a microcomputer so that the state of each pattern matching position of VS horizontal scanning lines of the horizontal candidate point detection circuit can be ascertained. If the output of each latch unit corresponding to the VS individual inspection coordinate positions of the horizontal candidate point detection circuit is larger than the threshold value in any one of the latch units corresponding to the individual inspection coordinate positions, if there is at least one latch, By maintaining the state of the horizontal scan line, the microcomputer is notified that there is a horizontal correlation line exceeding the correlation threshold value. That includes the candidate point detecting circuit, a pattern recognition device which is characterized in that the detected position coordinate position having the maximum value of the correlation data for a single or a plurality of candidate points.