JPH0442710B2 - - Google Patents

Description

[Detailed Description of the Invention] Industrial Application Field The present invention provides a method for quickly determining 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, for example. The present invention relates to a pattern matching method used in a two-dimensional position detection method.

BACKGROUND TECHNOLOGY Conventional position detection methods using pattern matching can be broadly classified as follows: 1. After storing one frame of an image pattern obtained from an imaging device in an image pattern storage device such as a semiconductor memory, matching processing with a standard pattern is performed. Method.

2. A device that generates the coordinates of the position being scanned by the imaging device, and an image pattern storage device such as a semiconductor memory that has the minimum capacity necessary for matching the standard pattern with the standard pattern. A method of performing matching processing while importing image patterns.

It can be classified into two types.

Therefore, we will examine three items regarding these two pattern matching methods: processing speed, processing extensibility, and cost.

First, regarding the processing speed, the first method performs the matching process after importing one frame of the image pattern into the storage device, which generally requires processing time and is not suitable for real-time processing. If the time required to import one frame of the image pattern into the storage device is T _i , and the time required to match the one frame worth of image pattern imported into the storage device with the standard pattern is T _p , then according to the first method, The time T _n1 required for pattern matching is expressed by the sum of T _i and T _p as follows in equation (1): T _n1 =T _i +T _p (1). The second method performs the matching process while capturing the image pattern, so the processing time is short and it is suitable for real-time processing. The time T _n2 required for pattern matching by the second method is
It is equal to T _i and is expressed as in equation (2).

T _n2 = T _i ...(2) In the first method, in order to speed up the process, matching reference points are set roughly every few pixels on the image pattern, and matching is performed for one screen first. A pattern matching method with a hierarchical structure is adopted in which recognition points are searched for by detailed matching using a search method based on matching reference points with a sufficiently high degree of matching obtained from the matching results. There are many.

Next, we will discuss the extensibility of processing. In the first method, the image pattern obtained from the imaging device is stored in the storage device, so the weighting of operators can be freely set in processing such as filtering, which is preprocessing for matching with standard patterns. After the matching process is completed, it is possible to selectively extract a specific area of the image pattern again and perform the confirmation process based on the result. Furthermore, when performing matching processing, it is also possible to scan the alignment point between the image pattern and the standard pattern (hereinafter referred to as matching reference point), which serves as a reference for matching, in a direction perpendicular to the scanning direction of the imaging device. . In the second method, preprocessing such as filtering and confirmation processing after matching processing are possible, but the preprocessing is fixed and the weights of operators during filtering processing cannot be easily converted. Also, the confirmation process after the matching process cannot be performed on the same screen where the matching result was obtained.

Finally, we will discuss the cost of realizing a pattern matching device using these two methods.
The first method requires more storage elements such as semiconductor IC memory than the second method in order to store one frame of image patterns, and the second method uses image patterns sent in real time as standard. Since this method performs pattern matching processing, it requires more logic ICs than the first method. In the past, semiconductor IC memories were expensive, making it difficult to realize the first method at a low cost.However, in recent years, the degree of integration of semiconductor IC memories has increased rapidly, and the price per unit storage capacity has decreased, making it difficult to realize the first method at low cost. Logic for realizing method 2
They are becoming cheaper than IC prices. Therefore, at present, it is possible to realize a pattern matching device at a lower cost by adopting the first method.

From the above, if the first method is adopted, a device with expandable functions that requires less processing time can be realized at low cost, while if the second method is adopted, the functions are limited and the cost is slightly lower. It can be said that it is possible to realize a device capable of high-speed processing, although it is expensive.

The present invention relates to a pattern matching method for performing two-dimensional position detection in real time using the first method described above. The reason why the first method, which is unsuitable for high-speed processing, was intentionally adopted is because the pattern matching process is highly scalable and flexible, as described above, and it can be realized at a lower cost than the second method. Further, the present invention is based on the premise of incorporating a high-speed processing technique called a hierarchical pattern matching method using the above-mentioned coarse matching process and detailed matching process.

Conventionally, in the hierarchical pattern matching method that aims to speed up pattern matching processing by combining coarse matching processing and detailed matching processing, all of the first matching reference points coarsely provided every few pixels on the image pattern are performs the first matching process with the standard pattern, and compares all of the resulting first matching degrees with the first passing standard value, which is a predetermined standard value, and determines whether the first matching degree is the first matching degree. The coordinates of the P first matching reference points whose values are equal to or higher than the acceptance standard value are retained, and the detailed information is re-examined for the small areas set in the vicinity of these P first matching reference points. A second matching process, which is a matching process, is performed, and a second matching reference point that gives the highest matching degree among the second matching degrees obtained as a result is set as a recognition point.

Problems to be Solved by the Invention However, with the above configuration, for example, it is difficult to detect the positions of multiple IC components at high speed and accurately within the processing time required when mounting IC components on a circuit board. That's difficult.

Figure 2 shows a standard pattern 31 with size M x M.
and the comparison reference point 32 of the standard pattern 31 and N×
An image pattern 33 having a size of N and first matching reference points 34 provided every n pixels on the image pattern 33 are shown. The first step is a rough matching process.
The matching process is based on the matching reference point 3 of the standard pattern 31.
2 to the first comparison reference point 34 of the image pattern 33, and the two patterns are compared. Conventionally, this first matching process is performed at all of the first matching reference points 34,
Thereafter, a second matching process, which is a detailed matching process, is performed for each of the first matching reference points that gave a degree of matching equal to or higher than the first acceptance criterion value in the first matching process. However, the number of first matching reference points is enormous, and the processing time required to match the standard pattern at all the first matching reference points alone is several times the time required to capture the image pattern. In the example of FIG. 2, the number of first matching reference points existing on the image pattern 33 is given by the following equation (3).

[N-M/n+1] ² ...(3) The symbol [a] is a Gauss symbol and represents the largest integer not exceeding a. For example, if N=512, M=16, and n=4, the number of first reference points becomes about 15,600.

Therefore, in order to shorten the time required for the first matching process at these many first matching reference points,
When the first verification process is passed, the second verification process is immediately performed to detect recognition points, and the process is terminated when a predetermined number of recognition points is obtained. However, as described above, if it is attempted to detect recognition points while mixing the first verification process and the second verification process, a new problem arises in that the position of the recognition point cannot be detected accurately.

FIG. 3 shows three first reference points 35, 36,
37. Second verification processing area 3 at each reference point
8, 39, 40, a matching reference point 41 where the degree of matching is maximum near the area shown in the figure, an area 42 where there is a matching reference point giving a degree of matching exceeding the second acceptance criterion value;
A matching reference point 43 with the second highest degree of matching in the area is shown. As is clear from the figure, the areas 38, 39, and 40 where the second matching process is performed are set so that adjacent areas slightly overlap each other. Three areas 38 for performing this second matching process,
When the comparison process with the standard pattern is performed in 39 and 40, the comparison reference point 43 is reached in both areas.
The degree of matching at is maximum. Verification reference point 43
is included in the area 42 where the matching reference point exists that gives a degree of matching that exceeds the second acceptance standard value. Matching reference point 4
3 is detected as a recognition point. However, in reality, a true recognition point with a higher degree of matching exists at the lower right of the matching reference point 43. Therefore, the recognition point 43 detected from the three first comparison reference points 35, 36, and 37 is an erroneous recognition point.

This is because the passing criteria for the second verification is set to have a range of over a certain standard value, so even if it is not the maximum value, there is a possibility that it will pass and become a recognition point. are doing.

In view of the above-mentioned problems, the present invention provides a hierarchical pattern matching method aimed at quickly and accurately detecting a plurality of recognition points existing in an image pattern.

Means for Solving the Problems In order to solve the above problems, the pattern matching method of the present invention stores an image obtained from an imaging means as a digitized image pattern in a storage means, and stores the image pattern as a digital image pattern. In a pattern matching process in which a characteristic partial area 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, the image pattern and the standard pattern are A matching reference point, which is a registration point of When the first matching degree obtained as a result of the first matching process becomes larger than the first reference value given in advance by performing the matching process, a second match is made near the first matching reference point. 2, a plurality of matching reference points are generated at a second interval narrower than the first interval, and a second matching process is performed with the standard pattern at all of the plurality of second matching reference points, The second matching reference point at which the matching degree is maximum among the plurality of second matching degrees obtained as a result of the second matching is
Among the areas where the plurality of second matching reference points are present, there are no matching reference points in the area facing the boundary with the area where the image pattern matching has not yet been performed, and the second matching reference points given in advance are By validating the second matching standard with the maximum matching degree only when the second matching standard becomes larger than the reference value No. 2, the recognition points existing on the image pattern can be identified. It has a configuration to detect.

Function First, for the sake of simplicity, the image pattern and the standard pattern are assumed to be binary patterns of "O" or "1", where "O" represents black and "1" represents white. This will be explained with reference to.

FIG. 4 shows a flowchart of a conventional hierarchical pattern matching method. First, a matching process 51 is performed on all the first matching reference points for a given image pattern, and then a passing point whose degree of matching is equal to or higher than the first passing reference value is detected 52 in the first matching process. Then, a second verification process 53 is performed on the detected passing score. As a result, if a recognition point is detected, the recognition point data is stored 54.
Of the four processes shown in FIG. 4, the one that requires the most processing time is the first matching process 51 at all first matching reference points. Moreover, with this processing method, the time required for the matching process 51 at all the first matching reference points is fixed and cannot be completely reduced. Therefore, in the hierarchical pattern matching method of the present invention, the first matching process is performed on all first matching processes based on the flowchart shown in FIG.
In order to speed up the processing, the matching reference points are not necessarily checked. In FIG. 5, first, a verification process 55 is performed at a first verification reference point, and a failure determination 56 of the first verification process is made by immediately comparing it with a first acceptance standard value. If the result is a failure, no processing is performed, and the last first one on the image pattern is
A determination 57 is made as to whether or not the first matching reference point has been reached, and this processing is continued until the processing for all the first matching reference points is completed. If the verification process 55 at the first verification reference point is passed, the first
A second matching process 58 is performed from the matching reference position, and the maximum value of the matching degree obtained in the second matching process is compared 59 with the second passing reference point. A determination 57 is made as to whether or not the reference point 1 has been reached. If the second verification process 58 is passed, the recognition point data is stored 60, and then a determination 61 is made as to whether or not a predetermined number of recognition points have been detected, and if so, all processing ends. However, if not, a determination 57 is made as to whether the final first reference point has been reached. In this way, in the present invention, when a predetermined number of recognition points are detected, the process is stopped at that point, thereby reducing the number of times of the first matching process and speeding up the process. The matching process at the first matching reference point provided every n pixels is a rather rough process for determining whether there is a possibility that a recognition point exists in the vicinity. It is not necessary to match the entire area of the standard pattern with the image pattern, and as shown in FIG.
It is sufficient to match a portion 63 of 2 with its corresponding image pattern. In this way, limiting the matching process at the first matching reference point to only the partial area 63 of the standard pattern 62 also helps speed up the process.

Next, in the present invention, in order to avoid misrecognition, if the matching reference point with the highest degree of matching obtained in the second matching process faces an area where no matching process has been performed, Considering that there is a possibility that a matching reference point showing an even higher matching degree will appear in the future, the matching degree at this matching reference point is invalidated. FIG. 7 is a diagram showing such an example, in which the first reference point 6
4 is passed, and the second verification process is performed on the small area 65. As a result, the degree of matching is maximum at the second matching reference point 66, but in reality, the matching degree is maximum at the matching reference point 67 near this point, and this matching reference point 67
is included in an area 69 in which second matching processing is performed, which is determined from the next first matching reference point 68. Therefore, as shown in FIG.
Corresponding to the directions in which υ _x 71 and υ _y 72 have moved,
In the small area 73 where the second verification is performed, an area 74 facing the unverified area is defined as an invalid area. Then, the matching reference point 75 that has the highest degree of matching in the area 73 for performing the second matching process determined from the first matching reference point 70 becomes a point on this invalid area 74, and this In such a case, the matching reference point 75 with the highest degree of matching is not selected. Then, as shown in FIG. 9, the matching reference point 78 that exists in the region 77 for performing the second matching determined from the next first matching reference point 76 and has the maximum matching degree is selected as the recognition point. This recognition point 78 is located next to the collation reference point 79 that almost became the erroneous recognition point earlier, and is not included in the invalid area 80 at this time. In this way, in the small area where the second matching is performed, the area facing the unmatched area is defined as an invalid area, and if there is a matching reference point within this area that has the highest degree of matching, this matching reference point is By not taking it up, we ensured accurate recognition.

Embodiment A pattern matching method according to an embodiment of the present invention will be described below with reference to the drawings. 1st
The figure shows a block diagram of a pattern matching method which is an embodiment of the present invention. In FIG. 1, a video signal 2 obtained from an imaging means 1 such as a television camera is binarized by a binarization means 3 to become an image pattern 4 and stored in an image pattern storage means 5. The image pattern storage means 5 is controlled by an image pattern control means 6, and the input/output of the image pattern is performed by commands 8 from the central processing means 7. On the other hand, the standard pattern 9 is cut out from the image pattern storage means 5 at the time of teaching and held in the standard pattern storage means 10 prior to the pattern matching process. The standard pattern storage means 10 is controlled by a standard pattern control means 11, and inputting and outputting of standard patterns is performed by commands 12 from the central processing means 7. The central processing means 7 also reads out pattern data 13 and 14 to be checked from the image pattern storage means 5 and the standard pattern storage means 10, respectively, when performing the matching process between the image pattern and the standard pattern.
It has the role of transmitting the command 8 to the image pattern control means 6 and the command 10 to the standard pattern control means so as to be input to the logical operation means 15, and the role of controlling each block in FIG. Two pattern data 13 input to the logic operation means 15,
14, the matching is performed by negating the exclusive OR, and if the corresponding pixels in the two pattern data match, the matching result is "1".
otherwise, it becomes "0". When the standard pattern is composed of M _x × M _y pixels, this logical operation is performed on M _x pixels at a time, and the M _x “0” and “1” obtained as a result of the logical operation are ``The data 16 is sent to the matching degree calculation means 17, where the number of ``1'' is calculated.

For a standard pattern composed of M _x × M _y pixels, the above process is repeated M _y times to complete the matching process for the entire standard pattern. coming
Cumulative addition is also performed while calculating the number of "1"s in the M _x "0"/"1" data 16. In the case of the first matching process, the calculated degree of coincidence 18 is sent to the central processing means 7, where it is compared with the first acceptance criterion value. In the case of the second matching process, for example, the second
When the area to be subjected to the matching process consists of 5 x 5 second matching reference points, the matching process with the standard pattern is performed continuously at 25 matching reference points, and the matching degree calculation means 17 performs the matching process with the standard pattern. Calculated match degree 19
are all sent to the maximum matching degree storage means 20, where they are compared with the maximum matching degree, and the maximum matching degree is stored. Following the match calculation, the comparison with the maximum match is 25
times, at that point the maximum matching degree storage means 2
The maximum matching degree 21 held at 0 is sent to the invalid area determining means 22. Invalid area determination means 22
Then, it is determined whether or not the matching reference point showing the maximum matching degree exists in the invalid area, and if it is included in the invalid area, the matching degree is reset to 0.
The maximum matching degree 23 obtained as a result of the second matching process is sent to the central processing means 7, where it is determined whether or not a sufficient matching degree has been obtained as a recognition point. And when it becomes a point of recognition,
Recognition point data 24 consisting of the degree of coincidence and the x and y coordinates of the comparison reference point is stored in the detected position storage means 25, and when all processing is completed, all detected positions stored in the detected position storage means 25 are stored. data 2
6 is sent to the upper controller 27.

An outline of the operation of the pattern matching method configured as described above will be explained below using FIG. 10. In FIG. 10, a standard pattern 90 having a size of M×M pixels has N×N pixels in advance.
A characteristic portion in an image pattern 91 having a pixel size is given. A first matching point 92 is set for every n pixels on the image pattern 91, and the first matching process starts at a matching reference point 93 which is the starting point.
The image moves from left to right in the horizontal direction toward the collation reference point 94, which is the end point, and progresses overall from top to bottom. In the matching process shown in FIG. 1, the standard pattern 90 and the image pattern 9 are
1 is compared, and by performing the comparison only on the partial area 95 of the standard pattern, the processing speed is increased. In FIG. 10, two recognition points 96 and 97 are indicated by cross symbols, and these recognition points are detected from first reference points 98 and 99, respectively. If the number of recognition points to be determined is two, the process ends when the recognition point 97 is detected from the first comparison reference point 99. Therefore, image pattern 91
For the first matching reference points 92 that exist in large numbers, the first matching process is performed only on about half of the first matching reference points that are filled in black. The time required for the matching process can be reduced to approximately half that when the entire screen is processed, and the pattern matching process is accelerated.

Next, using FIG. 11, a state in which a recognition point is detected after passing the first matching process is shown. A small area 101 consisting of 5×5 pixels is set around the matching reference point 110 that passed the first matching process. This area becomes the area where the second matching process is performed. In the second matching process, all these 25 points are
A comparison is made with a standard pattern 102 as a reference point for comparison. At this time, unlike the first matching process, the entire area of the standard pattern 102 becomes the matching target area. As a result of the second matching process, the maximum matching degree is obtained at the matching reference point 103, but since this matching reference point 103 exists outside the invalid area 104, the matching degree at this time becomes the maximum matching degree. , is compared with the second acceptance standard value and becomes a recognition point. When the second matching process is completed, the process starts again with the first matching process.
Returning to the matching process, the restart point of the first matching process at this time is the first matching reference point next to the previous first matching reference point 100.
becomes the reference point 105 for comparison.

As described above, according to this embodiment, in the first matching process, only partial areas of the standard pattern are matched, and if a predetermined number of recognition points can be detected, the process is interrupted at that point. This speeds up the pattern matching process, and in matching processes limited to small areas, by setting invalid areas, it is possible to accurately extract the maximum point of matching and reduce misrecognition. .

In addition, in the embodiment of the present invention, misrecognition is detected in two stages of the first matching process and the second matching process, but depending on the size of the image pattern/standard pattern and the complexity of the pattern to be recognized, Eventually, the first
The process may be performed in two or more stages by increasing the setting interval of the comparison reference points and performing several comparisons until the final comparison in which the comparison reference points are set for each pixel. FIG. 12 shows an example of three-stage matching. First matching reference point 11 provided every 4 pixels
Partial areas 111 of the standard pattern one after another for 0
It is assumed that the verification is performed at the verification reference position 112 and the result is passed. Next, the first verification reference point 1 that passed
13 as a reference, 3×3 second matching reference points 114 are set for every two pixels. For the nine matching reference points, matching is performed using the partial area 115 of the standard pattern. If the matching reference point 116 with the maximum matching degree as a result of nine matching processes is not included in the invalid area 117, this maximum matching degree is compared with the second acceptance standard value, and if it is greater than or equal to that value, then If you pass the exam, you will pass the exam.
In the third matching process, which is the final stage matching process,
3×3 matching reference points 119 are provided for each pixel so that the passing score 118 of the second matching process is centered. The nine matching reference points are matched with the partial area 120 of the standard pattern, and the matching reference point 121, which has the maximum degree of matching obtained as a result of nine matching processes, has the third highest degree of matching. Effects of the invention in which a recognition point becomes a recognition point when the value exceeds the acceptance standard value As described above, the present invention performs matching on a small area in an image pattern consisting of a plurality of matching reference points, and measures the degree of matching within the small area. When finding the matching reference point that has the maximum, the part where the small area touches the unmatched area of the image pattern is considered an invalid area, and if the matching reference point that shows the maximum degree of matching exists in this invalid area, that matching reference point is By disabling
The area where the degree of matching is maximum is extracted to enable accurate position detection. In addition, the size of the matching area of the standard pattern to be matched with the image pattern is determined by the interval at which the matching reference points are set, and when a predetermined number of recognition points are obtained, the point in time is determined. By interrupting and terminating the process, high-speed position recognition became possible.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of a pattern matching method that is an embodiment of the present invention. FIG. Figure 3 is a diagram showing the matching reference point, and Figure 3 is a diagram showing the area where the second matching process is performed, which is set based on the first matching reference point, as well as erroneous recognition points and true recognition points. FIG. 5 is a flowchart showing an algorithm for hierarchical pattern matching according to the present invention. FIG. 6 is a flowchart showing an algorithm for hierarchical pattern matching according to the present invention. FIG. FIG. 7 is a diagram showing the partial area, and FIG. 7 is a diagram showing that the matching reference point at which the degree of matching is maximum in the second matching process is not necessarily the maximum point. FIG. 8 is a diagram showing the second matching processing area and the invalid area. The figure shown,
FIG. 9 is a diagram showing that the matching reference point at which the degree of matching becomes maximum is detected in the second matching process;
Figure 0 is a diagram showing that the process is interrupted and terminated when a predetermined number of recognition points is obtained, Figure 11 is a diagram showing that recognition points are detected through two-stage matching, FIG. 12 is a diagram showing that recognition points are detected through three stages of matching. 5... Image pattern storage means, 7... Central processing means, 10... Standard pattern storage means, 15...
Logical operation means, 17... Matching degree calculation means, 20...
...Maximum matching degree storage means, 22...Invalid area determination means, 32...Standard pattern side matching reference point, 34...
...Image pattern side matching reference point, 41... Point to be recognized, 42... Area where a sufficiently high degree of matching can be obtained as a recognition point, 43... Erroneous recognition point, 63... Standard pattern partial area, 74... Invalid region.

Claims (1)

[Scope of Claims] 1. An image obtained from an imaging means is stored as a digitalized image pattern in a storage means, and a characteristic partial area of the image pattern is stored in advance as a standard pattern in a storage separate from the storage means. In a pattern matching process in which the image pattern and the standard pattern are matched, a first matching reference point, which is a positioning point between the image pattern and the standard pattern, is set on the image pattern. Set as the first reference point at the interval,
A first standard in which a first matching degree obtained as a result of the first matching process is given in advance by performing a first matching process between the image pattern and the standard pattern at the first matching reference point. If the value is larger than the value, a plurality of second reference points are generated near the first reference point at a second interval narrower than the first interval, and A second matching standard is obtained by performing a second matching process with the standard pattern at all of the second matching reference points, and having a maximum matching degree among a plurality of second matching degrees obtained as a result of the second matching process. The point does not exist in the area facing the boundary with the area where the image pattern matching has not yet been performed in the area where the plurality of second matching reference points exist, and Only when the second reference value is greater than the second reference value, the second
A pattern matching method characterized in that a plurality of recognition points existing on the image pattern are detected by validating a matching criterion. 2. The method according to claim 1, in which the image pattern and the standard pattern are matched only in a part of the standard pattern, if the set interval of the matching reference points is one pixel or more. Pattern matching method.