JPH0566636B2 - - Google Patents

Description

[Detailed Description of the Invention] Industrial Application Field The present invention is directed to, for example, detecting at high speed the position of a specific pattern present in an image pattern obtained from an imaging device, which is necessary when mounting IC components on a circuit board. The present invention relates to a pattern matching method for two-dimensional position detection.

BACKGROUND TECHNOLOGY Conventional position detection methods using pattern matching can be roughly divided into the following: 1. After an image pattern obtained from an imaging device is temporarily stored in a storage device, a central processing device such as a microcomputer is used to issue an instruction to the central processing device. A method of matching image patterns and standard patterns using a group of programs. 2. By passing the image pattern obtained from the imaging means through a circuit dedicated to pattern matching processing,
A method for matching image patterns and standard patterns.

It can be classified into two types.

Conventionally, in the field of production technology such as position detection of IC components, the second method was often used in terms of high-speed processing, and also in terms of cost, because semiconductor memory for storing image patterns was expensive. , the second method was advantageous. However, in recent years, the degree of integration of semiconductor memory has increased significantly, and the price per unit memory capacity has decreased year by year.
By adopting this method, it has become possible to realize a pattern matching device at a lower cost. In addition, in the second method, the image signal obtained from the imaging means is processed in real time, so if the image is captured with a standard television camera that uses the interlaced scanning method, the output of the television camera can be Since only half of the scanning signal can be used, it is not possible to easily increase the resolution of the digitized image pattern. However, in the first method, the scanning signal output from the television camera is converted into digital information and then stored in a storage means. The resolution of the image pattern can be easily increased because it can be effectively used by storing it in the image pattern. However, in the first method, the image pattern obtained from the imaging means is stored in a storage means such as a frame memory, and then pattern matching is performed by a program, so the processing speed is considerably faster than in the second method. Attempts have been made to speed up the processing by combining slow, coarse pattern matching and detailed pattern matching, but it cannot be said that the speed has been sufficiently increased compared to the second method.

Problems to be Solved by the Invention As described above, the first method described above has difficulty in high-speed processing, and the second method described above has problems in terms of resolution and cost. There was a problem in that the device could not be realized at low cost.

In view of the above problems, the present invention provides a pattern matching method for realizing a high-speed, high-resolution pattern matching device at low cost.

Means for Solving the Problems In order to solve the above problems, the pattern matching method of the present invention uses an image pattern obtained by digitizing an image signal obtained by imaging a recognition target object with an imaging means. A characteristic partial area of the image pattern is stored in a storage means as a standard pattern in advance in a storage means separate from the storage means, and a matching process between the image pattern and the standard pattern is performed. In the pattern matching process that detects the recognition point corresponding to the characteristic partial area where the matching degree is maximum, matching reference points that serve as the reference for alignment of the two patterns are successively set on the image pattern. The order of generation is determined in accordance with the arrangement of a plurality of characteristic partial areas in the image pattern.

Furthermore, in addition to the first invention, a second invention provides a positional relationship between adjacent characteristic partial areas when a plurality of characteristic partial areas in an image pattern are regularly arranged. is retained prior to normal recognition processing, and when a recognition point, which is a position representing a characteristic partial area, is detected for the first time in normal recognition processing, the adjacent characteristic The second recognition point can be detected at high speed by limiting the existence range of the second recognition point adjacent to the recognition point detected for the first time to a small area using the positional relationship between the partial areas. , if there is an n-th recognition point (n is a positive integer of 2 or more), calculate the positional relationship between the adjacent characteristic partial areas from the position where the n-1st recognition point exists. In this pattern matching method, the range in which the n recognition points exist is limited to a small area using the n-th recognition point, and the n-th recognition point is detected.

In addition to the first invention, the third invention also provides a method for normally determining the positional relationship between a plurality of characteristic partial areas in an image pattern when the plurality of characteristic partial areas are not arranged regularly. When a recognition point, which is a position representing a characteristic partial area, is detected for the first time in normal recognition processing, the plurality of characteristic partial areas previously held are stored prior to recognition processing. Detect multiple recognition points in the image pattern by limiting the range of recognition points corresponding to characteristic partial areas that have not yet been detected to a small area using the positional relationship between them. The present invention has a pattern matching method characterized in that:

Effect The effect of the above technical means will be explained with reference to the drawings.

Figure 2 shows the direction and order in which the reference points are scanned in the image pattern. A method in which the reference points are scanned in the x direction and moved in the y direction (hereinafter referred to as the y-x scanning method)21 ,22,2
3, 24, and methods 25, 26, 27, and 28 in which the reference point is scanned in the y direction and moved in the x direction (hereinafter referred to as the x-y scanning method).
It is classified into four types depending on the direction and the y-direction code.

Figure 3 shows an image pattern 31 of a lead of an IC component that is an object to be recognized, a standard pattern 32 obtained by cutting out a characteristic partial area of the image pattern, a reference point 33 for comparison on the standard pattern side, and three recognition points 34, 3.
5,36 is shown. This image pattern 31 is scanned for reference points using the same method as a commercially available image pickup tube camera (21 in Fig. 2), and the standard pattern 31 is scanned.
When the matching process is performed, recognition points 34, recognition points 36,
Three recognition points were detected in the order of recognition point 35, and after the recognition point was detected, the storage order of the detected recognition point data was changed according to the coordinates of each recognition point, and the recognition point data and lead were It becomes necessary to take action. However, if the x-y scanning method (for example, 25 in Fig. 2) is adopted as the scanning method for the comparison reference points, the recognition points will be detected in the order in which the leads of recognition point 34, recognition point 35, and recognition point 36 are lined up. This not only allows correspondence between recognition point data and leads without changing the storage order of recognition point data, but also stops processing when the required number of recognition points is obtained (for example, the third
When the recognition point coordinates for the second lead from the left in the figure are required, the pattern matching process can be completed in a short time by stopping the process when the recognition point 35 is obtained.
In this way, when multiple characteristic partial areas exist in an image pattern and are lined up in the x direction, x
-y scanning method, if they are lined up in the y direction, y
-By adopting the x-scanning method to generate a reference point for comparison and performing the comparison process, it is possible to directly obtain the recognition point associated with the lead of the IC component.
For the above-mentioned reasons, the pattern matching process can be speeded up. Furthermore, if the codes of the x and y directions, which are the scanning directions on the matching standard, are determined according to the position of a characteristic partial area in the image pattern, the recognition point can be detected at a position as close as possible to the scanning start position. This can also be used to speed up pattern matching processing.

Which scanning method to adopt among the eight types of scanning methods shown in FIG. 2 is automatically determined based on the coordinates of the recognition point obtained during the teaching process. If only one recognition point is obtained during the teaching process, the position closest to the recognition point among the four corners of the image pattern is selected as the starting point of the pattern matching process, and x-y, y-x Which scanning method is selected depends on the x and y coordinate values of the recognition point.
A method that can be detected quickly is adopted. If the recognition points obtained during the teaching process are adjusted, the recognition points arranged regularly are extracted by analyzing the x and y coordinate values of multiple recognition points, and the recognition points at that time are Depending on the arrangement, either the x-y or y-x scanning method is automatically determined.

FIG. 4 shows that when characteristic partial areas in an image pattern are regularly arranged, after the first recognition point 41 is detected, when detecting the second and subsequent recognition points 42 and 43, Detected recognition point 4
1 and 42 by the amount of movement stored at the time of teaching, and set small areas 46 and 47 so that the moved points 44 and 45 are the centers of the moved points 44 and 45, respectively, and This shows that the area in which the pattern matching process for detecting the subsequent recognition points 42 and 43 is performed is limited to these small areas 46 and 47, thereby speeding up the process.

FIG. 5 shows the initial recognition point 51 even when the characteristic partial areas in the image pattern are not arranged regularly.
After detection, in detecting the second and subsequent recognition points, point 5 is moved from the first recognition point 51 by the amount of movement 52, 53 that is pre-stored in teaching.
Small areas 56 and 57 are set so that points 4 and 55 are at the center, and the areas where pattern matching processing for detecting the second and subsequent recognition points is performed are limited to these small areas 46 and 47. This shows that the speed can be increased. However, in this figure, it is assumed that the scanning of the matching reference point on the image pattern is performed by the method shown in FIG. 26.

As a result of the above operations, a pattern matching device capable of performing pattern matching processing between an image pattern and a standard pattern at low cost and at high speed can be achieved using the central processing means and a program which is a group of instructions for the central processing means. It can be realized.

Embodiment A pattern matching method according to an embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 shows a block diagram of a pattern matching method in an embodiment of the present invention. 1st
In the figure, an image signal obtained from an imaging means 2 according to a command from a central processing means 1 (including a group of commands for the central processing means) is binarized by a binarization means 3 to form an image pattern. The image pattern is stored in the image pattern storage means 4. In the teaching process performed prior to the normal pattern matching process, the teaching processing means 5 cuts out a standard pattern, which is a characteristic partial area in the image pattern, from the image pattern storage means 4 based on a command from the central processing means. ,
The standard pattern is stored in the standard pattern storage means 6. In the pattern matching process at the time of teaching, the matching reference point scanning means 7 determines the matching reference points to be set on the image pattern in the order shown in FIG. The image pattern data is read out and compared with the standard pattern data read out from the standard pattern storage means 6 in the hierarchical pattern matching means 8. As a result of the matching, the matching reference points for which a sufficiently high degree of matching has been obtained are treated as recognition points, and their coordinate data (hereinafter referred to as recognition point data) is sent to the recognition point storage means 9 and stored therein. The central processing means 1 issues commands to the verification reference point scanning means 7 and the hierarchical pattern matching means 8, respectively, until the pattern matching means is completed for the entire image pattern. When the pattern matching process for the entire area of the image pattern is completed, the plurality of recognition point data held in the recognition point storage means 9 is sent to the recognition point analysis means 10, where the number and arrangement of recognition points are analyzed. The scanning method for the verification reference point to be generated is automatically determined, and the result is sent to the verification reference point scanning means 7. In addition, the recognition point analysis means 10
When it is determined that the recognition points are regularly arranged, the average values of the amounts of movement in the x and y directions of adjacent recognition points are calculated from the recognition point data and stored in the movement amount storage means 11. is maintained.

In normal pattern matching processing, an image signal obtained from the imaging means 2 is first binarized by the binary means 3 according to a command from the central processing means 1, and is converted into an image pattern by the image pattern storage means 4. will be remembered again. This image pattern is compared with the standard pattern held in the standard pattern storage means 6 by the hierarchical pattern matching means 8 during the teaching process. The hierarchical pattern matching means 8 roughly matches the first pattern every several pixels on the image pattern.
Set a matching reference point, perform matching at the first matching reference point, and if a sufficiently high degree of matching is obtained, set a second matching reference point every other pixel near the first matching reference point. If a plurality of matching points are provided and the matching is performed again at the plurality of second matching reference points, and the maximum value among the obtained plurality of matching degrees is greater than or equal to a predetermined reference value. Processing is performed to detect it as a recognition point. According to this hierarchical pattern matching means 8, it is possible to significantly reduce the number of times pixel pattern data and standard pattern data are matched, compared to a normal pattern matching means that sets matching reference points for every other pixel. Become.

Further, at this time, the setting order of the first matching reference points, which are set one after another at intervals of several pixels on the image pattern, is automatically determined by the recognition point analysis means 10 during the teaching process, One of the eight scanning methods shown in FIG. 2 is selected.

When the pattern matching process is performed as described above and the first recognition point is detected, the central processing means issues a command to the hierarchical pattern matching means 8 to store the detected recognition point data in the recognition point storage means. 9
At the same time, the recognition point data is sent to the small area generating means 12. The small area generation means 12 receives movement data in the x direction and y direction from the movement amount storage means 10, and adds the respective movement amounts to the x and y coordinates of the recognition point, which is the recognition point data, to obtain a second recognition point. Calculate the center position coordinates of the small area where exists. The size of this small area depends on the movement distance calculated from the amount of movement in the x and y directions, the dimensional accuracy of the recognition target, and the rotation angle of the recognition target with respect to the imaging means when the image of the recognition target is captured. It is calculated based on a predetermined calculation formula.
In detecting the second recognition point, the processing target area is limited to the small area obtained in this way, and the matching reference point scanning means 7 detects the matching reference point within this small area based on an instruction from the central processing means 1. The hierarchical pattern matching means 8 performs a matching process between the image pattern and the standard pattern at each matching reference point, thereby performing the matching process at high speed. The obtained second recognition point data is stored in the recognition point storage means 9 and the small area generation means 12 as well as the first recognition point data.
The third recognition point is detected using the same procedure as the second recognition point detection. In normal pattern matching processing, the newly set small area completely falls outside the window frame of the image pattern, or no recognition point was detected within the set small area, or the newly set small area The process ends when the number of recognition points obtained is obtained. After the pattern matching process is completed, the plurality of recognition point data held in the recognition point storage means 9 is sent to the host controller 13 as an output of the pattern matching process.

Next, how recognition points are detected at high speed by the pattern matching method of the present invention for an actual image pattern will be explained with reference to the drawings.

Figure 6 shows two types of standard patterns 61, 62,
Image pattern 6 obtained by imaging the leads of IC components
3. Recognition points 64, 65, and 66 detected for the standard pattern 63 and recognition points 67, 68, and 69 detected for the standard pattern 62 are shown. When performing pattern matching processing using the standard pattern 61, the recognition point analysis means finds that characteristic partial areas on the image pattern are lined up in the x direction during teaching processing, and the x-y scanning method is automatically activated. However, the scanning direction at this time is limited to scanning directions 26 or 28 in FIG.
The reason why the scanning directions 25 and 27 in FIG. 2 are not adopted is because it is not known how many leads of the imaged IC component are present on the image pattern. If the scanning direction of 26 or 28 in Fig. 2 is adopted, the recognition points are detected in the order of 66, 65, and 64, and these recognition points are the first, second, and third leads counting from the right of the IC component lead. Since this corresponds to the main part, it corresponds to the leads of IC parts counted from the right in the order in which recognition point data are stored. According to the 26 or 28 scanning method in FIG.
Detection of the th recognition point 64 is performed at high speed because a small area is set based on the moving distance between the recognition points held during the teaching process, and the pattern matching process is limited to this small area. . Furthermore, when performing pattern matching processing using the standard pattern 62, the y-x scanning method of 23 or 24 in Fig. 2 is automatically adopted in the teaching processing, and recognition points are detected in the order of 69, 68, and 67. . When only specific recognition point data is required instead of all recognition point data existing in an image pattern, pattern matching processing can be speeded up by terminating the process when that recognition point data is obtained. can do. For example, in Figure 6, when performing pattern matching processing with standard pattern 62, the scanning method from 23 to 24 in Figure 2 will be adopted, but if recognition point data for the second lead from the bottom is required. In such a case, the processing time can be shortened by terminating the process when the recognition point 68 is detected and not detecting the recognition point 67.

As described above, by automatically determining the direction and order in which matching reference points are generated one after another in an image pattern according to the position and arrangement of a plurality of characteristic partial areas in the image pattern, the characteristic When the subregions are arranged regularly,
Based on the amount of movement between two adjacent recognition points corresponding to this characteristic partial area that has been stored in advance, the processing area for determining the second and subsequent recognition points is limited to a small area. Even if the characteristic partial areas are not arranged regularly, the following steps can be performed based on the amount of movement between recognitions corresponding to multiple characteristic partial areas stored in advance. By limiting the region in which the recognition points exist to a small region, the processing speed is increased.Moreover, the pattern matching processing is performed by the central processing means and a program that is a group of instructions for the central processing means. It can be realized at low cost because it is done in advance.

Effects of the Invention As described above, the present invention determines the direction and order in which matching reference points are generated one after another in an image pattern, depending on the position and arrangement of a plurality of characteristic partial areas on the image pattern. As a result, recognition points are detected in a state that corresponds to characteristic partial areas existing in the image pattern, and there is no need to change the order in which the detected recognition point data is stored after the pattern matching process is completed. It can be processed. Furthermore, in the second and third inventions, from the recognition point coordinates obtained during numerical processing, the average value of the amount of movement between adjacent points of recognition points corresponding to characteristic partial areas is calculated and held. During the recognition process, after the first recognition point is detected, for the detection of the second and subsequent recognition points, the area where the next recognition point exists is limited to a small area based on the above movement amount, and the pattern is detected only within this small area. All that is required is a matching process. This makes it possible to perform high-speed processing even when performing pattern matching processing using the central processing means and a program that is a group of instructions for the central processing means, thereby creating a pattern matching device that has high resolution and is capable of high-speed processing. can be realized at low cost.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing the configuration of a pattern matching method that is an embodiment of the present invention, Fig. 2 is a diagram showing eight types of methods for scanning a matching reference point on an image pattern, and Fig. 3 is a diagram showing the configuration of a pattern matching method that is an embodiment of the present invention. Figure 4 shows that recognition points are not detected in the order in which characteristic partial areas are lined up depending on the reference point scanning method.
Figure 5 shows how to speed up the processing by limiting the processing area for subsequent recognition check detection to a small area. FIG. 6 is a diagram showing how the lead of an IC component, which is an object to be recognized, is actually detected as a recognition point. 1...Central processing means, 2...Imaging means, 4...
Image pattern storage means, 5...Teaching processing means, 6
... Standard pattern storage means, 7 ... Verification reference point scanning means, 8 ... Hierarchical pattern matching means,
10... Recognition point analysis means, 12... Small area generation means, 21, 22, 23, 24... y-x scanning method, 25, 26, 27, 28... x-y scanning method, 31... Image Pattern, 32... Standard pattern, 34, 35, 36... Recognition point, 41, 42,
43...Recognition point, 46,47...Small area to be processed for recognition, 52,53...Movement amount from the first recognition point, 56,57...Small area to be processed for recognition, 61,
62...Standard pattern, 64, 65, 66...6
Recognition points for 1, recognition points for 67, 68, 69...62.

Claims (1)

[Scope of Claims] 1. An image signal obtained by capturing an image of an electronic component having a lead using an image capturing device is stored in a storage device as a digitized image pattern, and a characteristic portion of the image pattern is stored in a storage device. The region is stored in advance as a standard pattern in a storage means different from the storage means, and the image pattern and the standard pattern are matched to correspond to the characteristic partial region where the degree of matching is maximum. In the pattern matching process for detecting recognition points, a matching reference point, which serves as a reference for alignment of the two patterns, is scanned from the outside where no lead exists toward the inside where the lead exists, parallel to the direction in which the lead extends. Pattern matching method. 2. An image signal obtained by imaging an electronic component having leads with an imaging means is stored in a storage means as a digitized image pattern, and a characteristic partial area of the image pattern is preliminarily set as a standard pattern. A pattern that is stored in a storage means separate from the storage means and that performs a matching process between the image pattern and the standard pattern to detect a recognition point corresponding to the characteristic partial area where the degree of matching is maximum. In the matching process, the order in which matching reference points, which serve as a reference for alignment of the two patterns, are generated one after another on the image pattern is determined based on the arrangement of the plurality of characteristic partial areas in the image pattern. This is a pattern matching method that determines the positional relationship between adjacent characteristic partial areas when a plurality of characteristic partial areas exist in an image pattern and are arranged regularly. The adjoining characteristic portions are stored in advance of normal recognition processing, and when a recognition point is detected for the first time, which is a position representing a characteristic partial area in normal recognition processing, the adjacent characteristic portions are stored in advance. Detecting the second recognition point at high speed by limiting the existence range of the second recognition point adjacent to the recognition point detected for the first time to a small area using the positional relationship between the areas, In addition, if there is an n-th recognition point (n is a positive integer of 2 or more), the positional relationship between the adjacent characteristic partial areas is used from the position where the n-1st recognition point exists. A pattern matching method for detecting the n-th recognition point by limiting the range in which the n-th recognition point exists to a small area. 3. An image signal obtained by imaging an electronic component having leads with an imaging means is stored in a storage means as a digitized image pattern, and a characteristic partial area of the image pattern is stored in advance as a standard pattern. A pattern that is stored in a storage means separate from the storage means and that performs a matching process between the image pattern and the standard pattern to detect a recognition point corresponding to the characteristic partial area where the degree of matching is maximum. In the matching process, the order in which matching reference points, which serve as a reference for alignment of the two patterns, are generated one after another on the image pattern is determined based on the arrangement of the plurality of characteristic partial areas in the image pattern. This is a pattern matching method that determines the correspondence, and when a plurality of characteristic partial areas in an image pattern are not arranged regularly, the positional relationship between the plurality of characteristic partial areas is determined. is stored prior to normal recognition processing, and when a recognition point, which is a position representing a characteristic partial area, is detected for the first time in normal recognition processing, the above-mentioned characteristic points are stored in advance. By using the positional relationship between partial areas to limit the existence range of recognition points corresponding to characteristic partial areas that have not yet been detected to a small area, multiple recognition points that exist in the image pattern can be detected. A pattern matching method characterized by detecting.