JPH05288520A - Pattern matching method - Google Patents

Abstract

(57) [Summary] [Purpose] To provide a method for performing pattern matching at high speed. [Structure] A master object 1 is observed by a camera 3, a high resolution master image pattern M1 and a low resolution master image pattern M2 are obtained, and these image data are registered in an image memory 12. Further, the inspection object 1 is observed by the camera 3, the low resolution image pattern A and the high resolution image pattern B are obtained, and these image data are input to the image memories 11 and 12. Then, the low-resolution master image pattern M1 and the low-resolution image pattern A are matched to obtain an area E in which the inspection object 1 is expected to exist, and then the area E is compared with the high-resolution master image pattern M2. The inspection target 1 is detected by accurately matching the high resolution image pattern B.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a pattern matching method, and more particularly, to a pattern matching method capable of increasing the matching speed to shorten the processing time.

[0002]

2. Description of the Related Art Pattern matching performed for position detection and shape recognition of an inspection object such as an electronic component mounted on a board is data of each pixel of an image of the inspection object captured by a camera. Is stored in the image memory as image data such as black-and-white binary data or multi-valued data, and the stored data and the data of each pixel of the master image pattern stored in the memory in advance are collated one by one. It was done by doing.

However, since the conventional means collates the data of each pixel of the image of the object to be inspected with the data of each pixel of the master image pattern one pixel at a time, it takes a lot of processing time. was there.

Therefore, the present applicant previously proposed a method for speeding up pattern matching (Japanese Patent Laid-Open No. 2-9).
0847). This data uses the data of a plurality of pixels of the image as unit data and shifts the unit data by unit pixel to store the data in the address of the image memory, thereby the data of each address of the reference pattern memory for each address. The processing time can be greatly shortened by making it possible to collate with

[0005]

However, the above-mentioned conventional means has a problem that it is difficult to greatly reduce the processing time.

Therefore, an object of the present invention is to provide a pattern matching method which can greatly reduce the processing time by a simple means.

[0007]

To this end, the present invention provides a low resolution master image pattern and a high resolution master image pattern by observing a master object with a camera and stores these image data in an image memory. Register with. Further, the inspection object is observed by a camera to obtain a low resolution image pattern and a high resolution image pattern, and these image data are input to the image memory. Then, by matching the low-resolution master image pattern and the low-resolution image pattern to obtain the area where the inspection object is expected to exist, the high-resolution master image pattern and the high-resolution image pattern are found for this area. The object to be inspected is accurately detected by precisely matching.

[0008]

According to the above construction, an area in which the inspection object is expected to exist is obtained in advance from the low resolution image pattern, and precise pattern matching can be performed in this area with the high resolution image pattern. , Pattern matching can be performed at high speed.

[0009]

Next, an embodiment of the present invention will be described by taking the position detection of an electronic component mounted on a board as an example.

FIG. 1 is an overall configuration diagram for performing pattern matching according to the present invention. An electronic component (hereinafter referred to as “chip”) 1 is mounted on a substrate 2, and this chip 1
Is observed by the upper camera 3 and its position is detected by a pattern matching method described later. First, the chip 1 is observed by the camera 3 to obtain a low-resolution image pattern A and a high-resolution image pattern B, and the image data of these image patterns A and B are input to an image memory (described later).
Reference numeral 1'is an image of the chip 1, and the image 1'is input to the image memory as image data such as binarized data and multi-valued data. Reference numeral 4 is a low resolution processing unit.

FIG. 2 shows a low resolution master image pattern M.
1 is shown. FIG. 3 shows a high resolution master image pattern M2. These master image patterns M1 and M2 are obtained by selecting a chip having a good shape from a large number of chips as a master object and observing this with a camera 3, and a second image memory (described later). ) 12 is registered.

The low resolution processing by the low resolution processing section 4 is a well-known means, and its processing method will be briefly described with reference to the model patterns of FIGS. 5 (a) and 5 (b). Figure 5
(A) is a multi-valued image of 0, 1, 2 obtained by performing multi-valued processing on the raw image captured by the camera 3. This image is divided into groups each having four pixels G as one unit, as indicated by a dashed-dotted line grid. Then, the average value of the values of the four images G in each group is calculated. For example, the values of the four pixels G in group 1 are 0, 1, 1, and the average value thereof is 1.
Similarly, in group 2, the numbers are 2, 2, 2, and 2, and the average value is 2.

FIG. 5B is a low-resolution multivalued image obtained by obtaining the average value as described above for all the groups. In this embodiment, the resolution is reduced to 1/2, and as a result, the total number of low-resolution pixels G shown in FIG. 5B is 1 of the total number of high-resolution pixels G shown in FIG. It is / 4. This means that the total number of matching pixels becomes 1/4 by pattern matching,
This alone means that the matching time can be shortened to 1/4.

FIG. 4 is a block diagram of the circuit configuration. First
The image memory 11 of is connected to the serial port memory 20. The serial port memory 20 is connected to the input unit 21, and the raw image data captured by the camera 3 is input to the serial port memory 20 as one line every time one line is input into the first image memory as high resolution image data. It is input in 11 at a time. And this image memory 11
The low-resolution image pattern A, the low-resolution master image pattern M1, and the high-resolution master image pattern obtained by this processing are read by the microprocessor 13 of the data stored in The image data of M2 is input to the second image memory 12.

The first image memory 11, the second image memory 12, and the microprocessor 13 are connected by a data bus 14. A counter 16 is connected to the data bus 14 via an input port 15. The counter 16 counts the number of data input to the first image memory 11. The input port 15 reads the value of the counter 16. The counter 16 includes a synchronization signal generation circuit 17
It is connected to the. The synchronization signal generation circuit 17 controls the timing of counting and the timing of capturing images.
The counter 16 is connected to the selector 18. The selector 18 controls the serial port memory 20. The selector 18 is connected to the microprocessor 13 by an address bus 19.

The present means is constructed as described above, and the pattern matching process will be described below. First, a chip with a good shape is selected as a master object from a large number of chips, and this chip is observed by the camera 3,
The low-resolution master image pattern M1 shown in FIG. 3 and the high-resolution master image pattern M2 shown in FIG. 3 are obtained, and each image data is registered in the second image memory 12.

Next, the chip 1 to be inspected is mounted on the camera 3
The image pattern B of high resolution shown in FIG. 1 is input to the first image memory 11, and the image pattern A of low resolution is input to the second image memory 12 by observing. Here, the processing for obtaining the low resolution image patterns M1 and A is performed in the low resolution processing unit 4 by the method shown in FIG. Of course, the microprocessor 13 can also perform low resolution processing.

Next, as shown in FIG. 1, the master image pattern M1 is scanned in the image pattern A in the direction of the arrow N1, and the image 1'in the image pattern A and the image 1 in the master image pattern M1 are scanned. Find the area E in which 'matches. Here, since the image pattern A and the master image pattern M1 have low resolution, the detection accuracy is low, and therefore the accurate position of the image 1'in the image pattern A cannot be detected, but the image 1'in the area E Is very likely to exist. Further, the low-resolution image patterns A and M1 have a small number of pixels (in the case of FIG. 5, the number of pixels is 1/4 as described above), and the number of matching times is also 1/4 of that when the resolution is not reduced. Therefore, the processing time required for matching is reduced to about 1/16.

Therefore, the area E is set in the image pattern B of FIG. 1, and the high resolution master image pattern M2 is scanned in the area E in the direction of the arrow N2 to obtain the image 1'of the image pattern B. And the image pattern M
The position of the image 1'in the image pattern B is detected by precisely detecting the position where the second image 1'accurately matches,
That is, the position of the chip 1 on the substrate 2 can be accurately detected.

The size of the area E can be greatly reduced (for example, about 1/100) than the size of the image pattern B. Moreover, since the matching time for low-resolution data is also reduced to about 1/16, the overall processing time is reduced to 1/10 or less of the conventional method.

In the above embodiment, the position detection of the inspection object is described as an example, but the shape of the inspection object can be recognized in a short time by the same method.

[0022]

As described above, according to the present invention, the pattern matching with the high resolution image pattern only needs to be performed for a narrow area previously obtained with the low resolution image pattern, so that the processing time can be greatly extended. Can be shortened to a high level and high matching accuracy can be maintained.

[Brief description of drawings]

FIG. 1 is an overall configuration diagram of a pattern matching method according to the present invention.

FIG. 2 is a low-resolution master image pattern diagram according to the present invention.

FIG. 3 is a high-resolution master image pattern diagram according to the present invention.

FIG. 4 is a block diagram of a circuit according to the present invention.

5A is an image diagram of a model pattern according to the present invention. FIG. 5B is an image diagram of a model pattern according to the present invention.

[Explanation of symbols]

 1 chip (inspection object) 3 camera 11 image memory 12 image memory A low resolution image pattern B high resolution image pattern E area M1 low resolution master image pattern M2 high resolution master image pattern

Claims (1)

[Claims] 1. A process of (i) observing a master object with a camera to obtain a low-resolution master image pattern and a high-resolution master image pattern, and registering these image data in an image memory, (ii) ) A process of observing the inspection object with a camera to obtain a low-resolution image pattern and a high-resolution image pattern and inputting these image data into an image memory, and (iii) registration in the process of (i) above. A process of matching the low-resolution master image pattern obtained with the low-resolution image pattern obtained in the process (ii) to obtain an area where the inspection object is expected to exist, and (iv) the above ( In the area obtained by the process of iii), the high-resolution master image pattern registered in the process of (i) and the process of (ii) A pattern matching method comprising a process of matching the obtained high-resolution image pattern to accurately detect the inspection object.