JPH04188379A - Picture pattern recognizing method - Google Patents

Abstract

PURPOSE:To satisfactorily recognize a picture pattern at high speed by drawing a support line in the picture pattern at the read part of a recognition object equipment and detecting the position only by extracting the contour line of the picture pattern and detecting the intersection of the support line and the picture pattern. CONSTITUTION:A support line 11 is drawn while crossing the picture pattern corresponding to the read part of the recognition object equipment and next, the support line 11 is considered as one part of a contour line 13 in the picture pattern. Then, the contour line 13 of those synthesized picture patterns is extracted, intersections 15 of the support line 11 and the picture pattern are detected, and the position of the read of the recognition object equipment is decided from the coordinates of those intersections 15. Thus, the picture pattern can be recognized at high speed without being affected by the inclination of an input picture or the shape of the recognition object equipment.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to an image pattern recognition method for detecting the position and tilt of an electronic component as a workpiece in an electronic component mounting machine equipped with an image processing device. .

Conventional technology Conventionally, an electronic component mounting machine equipped with an image processing device performs QFP and PL based on images obtained from a visual sensor such as a television camera.
As a method for detecting the position of leaded electronic components such as CCs and TABs by image recognition processing, methods using template matching or partial corner detection of a pattern outline are generally proposed.

In the method using template matching, as a preparatory step, image pattern regions of characteristic parts such as lead tips of the target part are stored as a template image from a reference image taken of the recognition target part, and then in the recognition processing step, the image pattern area is stored first. A region matching the template image is searched on the input image, and the lead tip position is determined in detail based on the detected position and output.

On the other hand, in a method using a partial corner detection method of a pattern contour, the contour of the image pattern corresponding to the part to be recognized is extracted from the input image, and the contour line is adjusted to 180 degrees based on the curvature at each point of the contour. °Detect the folding point and output it as the lead tip position of the part to be recognized.

Problems to be Solved by the Invention When using the image recognition method according to the conventional technology described above, the method using template matching requires image comparison in order to search the input image for an area that matches the template image in the recognition processing step. processing is required,
It takes a lot of calculation time. Therefore, in application fields that require high speed, it is necessary to prepare special image processing hardware to increase the speed. In addition, it is necessary to capture a reference image and collect and store a template image in advance, and furthermore, if the difference between the reference image and the input image during recognition becomes large due to the influence of image tilt etc., areas matching the template image may not be detected. It may disappear and become unrecognizable.

On the other hand, in the method using corner detection, it is necessary to specify many parameters in order to detect the 180° corner of the outline of the image pattern. When the curvature of the contour line takes various values, it is difficult to obtain parameters to simultaneously detect those corners, so the scope of application is limited to electronic components with uniform lead width.

An object of the present invention is to overcome the drawbacks of the above-mentioned conventional techniques and to provide an image pattern recognition method that is fast and unaffected by the inclination of an input image or the shape of a part to be recognized.

Means for Solving the Problems The present invention first draws a support line that crosses an image pattern corresponding to a lead portion of a part to be recognized, and then assumes that the support line is a part of the outline of the image pattern. It is characterized by a process of extracting the contours of these composite image patterns, detecting intersections between the support line and the image pattern on the obtained contours, and determining the position of the recognition target component board from the coordinates of these intersections. do.

According to the above-described configuration, the present invention draws a support line on the image pattern of the lead part of the recognition target part, thereby extracting the outline of the image pattern and detecting the intersection between the support line and the image pattern at high speed and the shape of the recognition target part. The position of the recognition target part can be detected only by processing that is not affected by the

Embodiment As an embodiment of the present invention, a method for detecting the lead tip position of an electronic component with a lead, which is one of the target components in an electronic component mounting facility, will be described. In Figure 1, Snap 1 is at the point where the image is captured, and reflected light illumination 2 is shown in Figure 2.
In this step, the electronic component 4 with leads is imaged by the television camera 3 under the following conditions, and an input image 5 shown in FIG. 3 is obtained. 6 in the input image 5 is an image pattern of a lead portion of the electronic component 4 with leads. As shown in FIG. 3, when the reflected light illumination 2 is used, only the metal lead portion reflects light well, resulting in an input image dotted with isolated patterns of high brightness.

Step 2 is an image pattern inclination detection step, in which the rough inclination of the part to be recognized is detected by the following procedure. Third
As shown in the figure, a window 8 is designated on the image pattern corresponding to the lead portion of the part to be recognized so as to surround them. Next, the pixels within the window 8 are scanned in the longitudinal direction of the lead to detect a point 9 of change in brightness. The two-dimensional spread direction of the detected brightness change points is determined by the least squares method, and is determined as the inclination of the part to be recognized.

Step 3 is a support line drawing step, in which a support line 11 shown in FIG. 4 is drawn so as to cross the image pattern of the lead portion along the inclination of the recognition target part determined in the image pattern inclination detection step 2. . For the brightness of the support line 11, a brightness that cannot be obtained by each pixel of the input image is used so that it can be distinguished from the image pattern 6. For example, if the range of brightness of each pixel of the input image is limited to 0 to 127, a value such as 255 is set as the brightness of the support line.

Step 4 is a pattern contour extraction step, in which a composite image pattern of the support line 11 obtained in the support line drawing step 3 and the image pattern 6 is considered, and its contour line 1 is
Extract 3. The support line 11 is treated as a part of the image pattern to be detected, and a continuous contour line 13 is extracted. Note that by considering the support line 11 as a part of the image pattern as described above, multiple patterns can be isolated and scattered within the input image 5, like the lead image pattern 6 obtained in the image input process. Even if
A support line 11 connects those isolated patterns and 1
Since it can be handled as one image pattern, processing can be simplified and speeded up.

Step 5 is an intersection detection step, in which the outline 13 of the composite image pattern obtained in the pattern outline extraction step is scanned, and the intersection 15 between the support line 11 and the image pattern 6, that is, from the support line 11 to the image pattern 6, is scanned. The point where the image pattern 6 switches to the support line 11 and the point where the image pattern 6 switches to the support line 11 are detected and recorded. Whether the current scanning point is the intersection 15 is determined based on the brightness of the scanning point. For example, according to the example given in the support line drawing process above, the brightness of the support line 11 is 25
5 and the other image patterns 6 have a luminance of 0 to 127, so if the outline 13 is scanned while reading the luminance of the pixels at each point, the intersection 15 can be easily detected. Note that if the part to be recognized is an electronic component with leads, two intersections 15 occur for each lead, so two consecutive intersections 15 are generated for each lead.
Post-process processing of recording two intersections in pairs can be simplified.

Step 6 is a lead position detection step in which the lead tip position of the recognition target component is accurately determined based on the coordinates of the intersection 15 detected in the intersection detection step 5.

In the above processing procedure, the pattern contour extraction step 12 and the intersection detection step 14 do not need to be separate processes, and it is also possible to detect and record intersections during the contour extraction process.

Effects of the Invention As is clear from the above description, the present invention realizes an image pattern recognition method that is fast and unaffected by the inclination of an input image or the shape of a part to be recognized.

[Brief explanation of drawings]

Fig. 1 is a flowchart of the image pattern recognition method of the present invention, Fig. 2 is an explanatory diagram of the image input process, Figs.
The figure is an explanatory diagram of an image pattern recognition method to which the present invention is applied. 4...Electronic component with leads, 5...Input image, 11...Support line, 13...
...Contour line, 15...Intersection. Agent's name: Agent Haruaki Koebi and 2 others Figure 1 Start picture 1 Ikei Crab Step 6, □. , -52 Inclined image upper process, j-5 3 3- te V (turtle 5 2 figure drawing process Barry') Kanyp Kishu 4 extraction process

Claims (1)

[Claims] A first step of detecting an image pattern corresponding to the recognition target from an image of the recognition target and drawing an appropriate support line on the image pattern; A second step of extracting contour lines from a composite image pattern of the support lines and the image pattern, and each part of the contour line obtained in the second step belongs to the support line. 2 of the part and the part belonging to the outline of the image pattern
A third step is to classify the object into two types of contour segments, detect and record all intersections that are the junctions between these two types of contour segments, and a third step to determine the position of the recognition target based on the coordinates of the obtained intersections. Image pattern recognition method with four steps.