JPH0727563B2 - Pattern detection method - Google Patents

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to position detection drawn on a circuit board for detecting the position of the circuit board, which is necessary when mounting IC parts on the circuit board, for example. Detects the position of a figure at high speed, 2
The present invention relates to a pattern detection method for detecting a dimensional position.

Conventional technology When mounting IC components on a circuit board, detect the opposite positions of the IC component and the circuit board, correct the position / rotation angle of the IC component or the board, and then mount the IC component on the circuit board. Method.

Conventionally, the pattern matching method has been used to detect the positions of the IC component and the circuit board. In the pattern matching method, a characteristic part of a recognition target pattern is registered as a dictionary pattern at the time of teaching, and matching is performed while moving the dictionary patterns one after another on the image data obtained by the image pickup means at the time of recognition processing. The position of the dictionary pattern with the highest degree is detected.

Problems to be solved by the invention When the position detection method by pattern matching is used,
In the position detection of the component, there is little erroneous recognition and highly accurate recognition is possible, but some problems occur in the position detection of the circuit board. The reason for this is that IC parts have a characteristic part that can be registered as a dictionary pattern, such as the tip of a lead, and process high-contrast images captured by transillumination. However, in the position detection of the circuit board, the figures drawn on the circuit board for position detection are often circular and there are few characteristic parts that can be registered as a dictionary pattern. This is because such an image with a not very high contrast is processed, so that problems such as erroneous recognition and deterioration of recognition accuracy are likely to occur. Also,
Since the image itself is complicated, it takes a lot of time to remove noise. For this reason, attempts have been made to take countermeasures by adopting a pattern having a characteristic corner portion such as a square or an equilateral triangle in the figure for detecting the position of the circuit board. By its nature, figures are made by etching copper foil with the same plate as the circuit pattern, so the outlines of squares and equilateral triangles, especially the corners registered as a dictionary pattern, are likely to collapse, causing erroneous recognition and poor recognition. Therefore, it cannot be said that the above measures have always solved the problem.

Therefore, the above-described configuration has a problem that it is difficult to remove noise efficiently and to erroneously recognize the position detection graphic provided on the circuit board and to detect the position accurately. Was there.

Means for Solving the Problems In order to solve the above problems, the pattern detection method of the present invention sets a plurality of linear trajectories on image data, and sequentially reads and reads image data along the linear trajectories. The first step is to detect the position on the inner side of the pattern where the image data changes from a value representing the outside of the pattern to a value representing the inside of the pattern, and remove the other parts as noise. Furthermore, the pixel data in the vicinity of 8 is read out at each position on the linear trajectory existing after the position detected inside the pattern next, and all the pixel data in the vicinity of 8 have a value representing the inside of the pattern for the first time. By using the position as the representative position of the pattern, the representative position of the pattern is detected at high speed while removing a small noise pattern.

Effect of the Invention With the above-described structure, the present invention first removes, as a noise, a small pattern that does not intersect a predetermined linear trajectory from among the patterns existing on the image data, and then reads the image data along the linear trajectory. The position on the inner side of the pattern of the pattern intersecting the straight line trajectory is detected one after another, and the pixel data of 8 neighborhoods is read out at each position on the inner side of the pattern. By making the position having the value the representative position of the pattern, the pattern can be detected at high speed while removing noise.

Example Hereinafter, regarding the pattern detection method of one example of the present invention,
A description will be given with reference to the drawings.

FIG. 1 shows a block diagram of an embodiment of a pattern detecting method which is an embodiment of the present invention. In FIG. 1, the image signal obtained from the image pickup means 1 is binarized by the binarization means 2 and stored in the image data storage means 3. Next, the linear trajectory search means 4 sequentially reads the image data along one straight line from the image data storage means 3, and the image data changes from a value representing the outside of the pattern to a value representing the inside of the pattern. Detect the position of. And
At each position on the linear trajectory existing after the detected position on the inner side of the pattern, pixel data of 8 neighborhoods (hereinafter,
8 neighborhood data) The 8 neighborhood data of the image data is read by the reading means 5, and the position on the linear trajectory when all the 8 neighborhood data have a value representing the inside of the pattern for the first time is stored as the pattern representative position. It is stored in the means 6. By performing the above processing on a plurality of linear trajectories set on the image data, the representative position of the pattern is stored while removing a small noise pattern from the plurality of patterns existing in the image data. 6 can be stored one after another.

In the embodiment of FIG. 1, a method of obtaining the contour points of the pattern from the representative position of the pattern is used to obtain the finally required reference position of the pattern. The method for obtaining the pattern reference position from the pattern representative position will be described below in the first section.
Description will be given with reference to the drawings.

The contour detecting means 7 takes out the pattern representative positions stored in the pattern representative position storing means 6 one by one by the above-mentioned processing, and searches the pattern representative positions in the fixed direction starting from the taken pattern representative positions to obtain the contour points of the pattern. Is first detected. As a search method, the 8-neighborhood data reading means 5 reads the 8-neighborhood data, and the contents of the 8-neighborhood data are discriminated to know whether or not the contour point of the pattern has been reached. The contour point can always be detected by continuing to advance in the given direction until the contour point of the pattern is reached. After the contour points are detected, the direction in which the contour points are connected is calculated based on the 8 neighborhood data, and the contour points are detected one by one until the pattern makes one round, and the coordinate values of the contour points and the number of contour points are calculated as geometric features. It is stored in the quantity storage means. The geometric feature amount other than the above is the area of the pattern, and the pattern feature area is detected by the geometric feature amount detecting means 9 and stored in the geometric feature amount storing means 8. When the geometric feature quantity is detected for the patterns corresponding to all the representative positions stored in the pattern representative position storage means 6, the pattern evaluation means 10 reads out the contour point number / area from the geometric feature quantity storage means 8, The pattern is evaluated according to a predetermined evaluation formula, and the pattern closest to the taught pattern is selected. Then, the center of gravity of the selected pattern, the center of the rectangle circumscribing the pattern, the x, y of all contour points of the pattern
One of the average values of the respective coordinates (hereinafter referred to as the center of gravity of the contour) is calculated by the reference position detecting means 11 and output to the host controller 12.

Next, how a plurality of patterns existing in the image data are detected by the direct trajectory search and the reading of the 8-neighbor data will be described with reference to FIGS. 2 and 3.

FIG. 2 shows a plurality of patterns existing in the image data and a plurality of linear trajectories set on the image data. In FIG. 2, the image data 20 has four patterns.
It includes 21 ・ 22 ・ 23 ・ 24, while four linear tracks 25 ・ 26 ・ 27 ・ 28 are set. The linear trajectory 25 is
The pattern 23 is detected, and the linear trajectory 26 detects the pattern 22 and the pattern 24. The linear trajectory 27 again detects the pattern 22 detected by the linear trajectory 26, but the linear trajectory 28 does not detect any pattern. Further, the pattern 21 is not detected from any of the linear trajectories and is removed as noise.

FIG. 3 shows the positional relationship between the linear trajectory and the pattern. The linear trajectory 30 detects three patterns 31, 32, and 33, but two of the three patterns are removed as noise, and the final pattern. Only the pattern 32 is detected. First, when the image data is sequentially read out along the linear trajectory 30,
It can be seen that position 34 had a value representing the outside of pattern 31, whereas position 35 had a value representing the interior of pattern 31. That is, it is expected that the positions on the linear trajectory 30 existing after the position 35 are present inside the pattern 31. Therefore, the position on the linear trajectory after position 35 is pattern 31.
While existing inside (section from position 35 to position 36)
The image data of 8 neighborhoods is read at each position, and the position where all the values of the 8 neighborhood data have the values that represent the inside of the pattern 31 for the first time is held as the representative position of the pattern 31, but between the positions 35 and 36 , There is no position where the values of all the 8 neighborhood data have values representing the inside of the pattern 31, and the pattern 31 is removed as noise. The search on the linear trajectory 30 is performed once outside the pattern 31 and then again at the position 37.
However, all the values of the 8-neighboring data have values representing the inside of the pattern 32 only at the position 39 in the section from the position 37 to the position 38 which is the inner section of the pattern 32. As a result, the x, y coordinates of the position 39 are held in the storage means as a position representing the pattern 32. Furthermore, the search on the linear trajectory 30 enters the inside of the pattern 33 in the section from the position 40 to the position 41, and even in this section, there is a position where all the values of the 8 neighborhood data have values representing the inside of the pattern 33. The pattern 33 is also removed as noise like the pattern 31. It seems unreasonable that a large pattern like pattern 33 is removed as noise regardless of a small pattern like pattern 31. Must be detected by.

As described above, according to the present embodiment, a plurality of linear trajectories are set on the image data, the image data are read one after another along the linear trajectories, and the value of the read image data represents a value outside the pattern. The position inside the pattern that changes to a value representing the inside of the pattern is first detected, and then the 8 neighborhood data is sequentially read from the image data at each position of the linear trajectory existing after the detected position inside the pattern. , A plurality of patterns existing in the image data are detected at high speed while removing a small noise pattern by setting a position having a value representing the inside of the pattern for all of the 8 neighborhood data for the first time as a representative position of the pattern. be able to.

Further, a means for detecting all contour points of the pattern with the representative position of the pattern as a starting point, a means for detecting geometric feature amount of the pattern such as the number of contour points and an area, and a pattern closest to the taught pattern among several patterns , By adding a means for detecting the reference position of the pattern such as a pattern evaluation means for selecting the geometric feature amount of the pattern, a center of gravity of the pattern, a center of a rectangle circumscribing the pattern, and a center of gravity of the contour of the pattern, The pattern closest to the taught pattern can be selected from a plurality of patterns in the image data, the reference position thereof can be accurately detected at high speed, and output to the host controller.

Although the linear trajectory is set radially on the image data in the embodiment, the linear trajectory may be set as a group of horizontal straight lines on the image data.

As described above, according to the present invention, a plurality of linear trajectories are set on image data, a small amount of image data called image data on the linear trajectories is read out, and after the linear trajectories enter the pattern, , A plurality of patterns existing in the image data are obtained by reading the pixel data of the eight neighborhoods on the straight line trajectory, and setting the position when all of the eight neighborhood data exist inside the pattern for the first time as the representative position of the pattern. It has become possible to detect at high speed.

[Brief description of drawings]

FIG. 1 is a block diagram of a pattern detection method according to an embodiment of the present invention, FIG. 2 is a diagram showing a plurality of linear trajectories set for a plurality of patterns existing in image data, and FIG. 3 is a straight line. It is a figure which shows that a pattern is detected by the search on a track | orbit and a noise pattern is removed. 1 ... Imaging means, 2 ... Binarization means, 3 ... Image data storage means, 4 ... Linear trajectory search means, 5 ... 8 Neighborhood data reading means, 6 ... Pattern representative position storage means, 20
…… Image data, 21 …… Noise pattern, 22 ・ 23 ・ 24…
… Pattern, 25 ・ 26 ・ 27 ・ 28 …… Linear trajectory, 30 …… Linear trajectory, 31 ・ 33 …… Noise pattern, 32 …… Pattern, 39
...... Pattern representative position.

 ─────────────────────────────────────────────────── ─── Continuation of front page (72) Inventor Zenichi Okahashi 1006 Kadoma, Kadoma City, Osaka Prefecture Matsushita Electric Industrial Co., Ltd. (56) References JP-A-59-114682 (JP, A) JP-A-59-114682 211167 (JP, A) JP-A-59-136878 (JP, A)

Claims (1)

[Claims] 1. An image signal obtained by an image pickup means is once stored as digitized image data, a plurality of linear trajectories are set in arbitrary directions on the image data, and the linear trajectory is set along the linear trajectory. The image data is read one after another, the position of the inside of the pattern where the value of the read image data changes from the value representing the outside of the pattern to the value representing the inside of the pattern is detected first, and then the detected inside of the pattern is detected. Pixel data in the 8 neighborhoods are sequentially read from the image data at each position on the linear trajectory existing after the position, and a position where all the pixel data in the 8 neighborhoods have a value representing the inside of the pattern for the first time is the representative position of the pattern. Pattern detection method.