JPH04238206A - Multi-visual field recognizing method for pattern - Google Patents

Abstract

PURPOSE:To recognize the position of a part with size not contained in recognition visual fields via the recognition information from pickup images in the recognition visual fields set at multiple positions. CONSTITUTION:Recognition visual fields each serving as a unit pickup object for one corner of an angular pattern which is a recognition object and two sides of this corner are set at multiple corners of the angular pattern, and the required position of the angular pattern is determined from the position information of the corner of the angular pattern of the pickup image in each recognition visual field and the relative position information of the recognition visual fields.

Description

[Detailed description of the invention]

0001

[Field of Industrial Application] The present invention relates to a multi-view recognition method for a rectangular pattern, and more particularly to a multi-view recognition method for a rectangular pattern that recognizes a desired position from images captured in a plurality of recognition fields for one rectangular pattern. It is something.

0002

[Prior Art] Conventionally, when performing recognition processing in a method of finding the representative point of a rectangular IC component, often the center of the component, a part of the lead image of the component is used to find the representative point of the entire component. Alternatively, there is a method of recognizing the lead image of the entire part and finding the representative point of the part from the average value. Another method is to find an approximate straight line of the part outline and find the center of a circumscribed rectangle surrounded by this approximate straight line.

0003

[Problem to be Solved by the Invention] In the above method of obtaining information from a part of the lead image, even if the size of the part is larger than the field of view, it is only necessary to recognize a part of the part. There are no restrictions on the size of parts. However, it is difficult to get an overall picture of the part, and it is difficult to determine the true representative points. Furthermore, in the method of detecting leads of the entire component, it is necessary to image the entire component, and the size of the component is limited to the size of the field of view.

[0004] Furthermore, in the case of parts larger than the field of view, there is a method of detecting leads while moving the recognition field of view.
Conventional methods can only handle cases where there are leads on two opposing sides or on all sides. For parts that do not have leads or for which leads are difficult to distinguish, drawing a circumscribed rectangle from approximate straight lines on each side and finding representative points from it is limited to parts that can be photographed within one field of view.

Therefore, the present invention recognizes the position corresponding to the entire image from the captured image in the recognition field corresponding to a plurality of parts of one rectangular pattern, and determines the necessary positions such as representative points regardless of the size of the pattern. The object of the present invention is to provide a multi-view recognition method for rectangular patterns that requires high accuracy.

[0006]

[Means for Solving the Problems] In order to solve the above-mentioned problems, the present invention provides a recognition field of view in which one corner of a rectangular pattern to be recognized is a unit imaging target, and a recognition field of view in which one corner of a rectangular pattern to be recognized is divided into multiple corners of the rectangular pattern. The required position of the rectangular pattern is determined from the corner position information on the recognition field of view of the rectangular pattern image set corresponding to and captured in each recognition field of view, and the relative position information of each recognition field of view that recognized the position information. It is characterized by calculating and recognizing.

[0007]

[Operation] According to the above configuration of the present invention, a plurality of corners of a rectangular pattern are imaged in recognition fields of view set at different positions, but a plurality of corners of an image of a rectangular pattern in each recognition field of view are imaged. The positional information of , and the positional information of each recognition field of view that recognizes that positional information are the elements that uniquely determine the corner position in the overall image of the rectangular pattern. The corresponding required position can be calculated and recognized with high precision.

[0008]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below in detail with reference to the drawings.

This embodiment shows a case where the position of a large chip component 9, which is an electronic component having a rectangular planar shape as shown in FIG. 4 and subsequent figures, and the center position G as its representative point are recognized. Recognition is performed using a video camera 1 shown in FIG. The video camera 1 is controlled by a drive device 3 controlled by a microcomputer 2 to perform necessary operational controls such as imaging and changing the position of the recognition field of view.

Images captured by the video camera 1 are subjected to necessary recognition and processing by the microcomputer 2 via the image processing device 4. For these controls, the microcomputer 2 stores a preset program 5 and receives necessary external inputs from an input device 6.

The recognition operation procedure is as shown in the flowcharts of FIGS. 2 and 3. This will be explained.

First, the dimensions of the part 9 to be recognized in the X and Y directions are input using the input device 6, and necessary data of the part 9 is registered (step #1). Next, as shown in FIG. 4, the recognition field 7 of the video camera 1 is moved to a first position corresponding to the first corner 11 of the component 9, and an image is taken (step #2). Here, the first corner 1 in the recognition field of view 7
The positions x1 and y1 of the first corner 11 and the inclinations θx1 and θy1 of the two sides 21 and 22 forming the first corner 11 with respect to the X and Y axes are calculated and recognized, and these are registered (Steps #3 to #5
).

When this is completed, the recognition field 7 is moved to the position shown in FIG. 5 corresponding to the second corner 12 of the part 9 (
Step #6). This movement is performed by a predetermined amount based on the input data in step #1. In parallel to this, the predicted position of the second corner 12 is calculated from the input data at step #1 and the registered data at step #5 (step #7).

After waiting for the movement in step #6 and the calculation in step #7 to be completed (steps #8 and #9), imaging is performed using the recognition field of view 17 at the position shown in FIG.
10). Then, the positions x2 and y2 of the second corner 12 are calculated and recognized (step #11). Next, this position x
2, y2 is compared with the predicted position, and if it exceeds the allowable range, the part 9 is determined to be defective, such as a different part or a missing part, and subsequent detection operations are stopped (steps #12 and #13).

[0015] Furthermore, when this detection is performed on the mounted components in a component mounting machine, it is preferable to also stop the mounting operation.

If a defect is to be detected in step #12, after calculating and recognizing the inclinations θx2 and θy2 of the two sides 22 and 23 forming the second corner 12, the position x2 is detected.
, y2 and the slopes θx2 and θy2 (steps #14 and #15).

Next, it is determined whether the final screen corresponds to the fourth corner 14 in FIG. 7 (step #16). If it is not the final screen, return to step #6, and repeat the position x3 of the third corner 13, fourth corner 14 until the final screen is reached.
y3, x4, y4 and the slopes θx3, θy3, θx4, θy4 of the two sides 23, 24 and 24, 21 forming them are recognized by calculation and registered (steps #6 to #1
6).

If it is the final screen in step #16, the process moves to step #17, and each of the first to fourth corners 11 to
14 positions x1, y1 to x4, y4 and the two sides 21, 22, 22, 23 forming those corners 11 to 14,
23, 24, 24, 21 slope θx1, θy1 to θx4
, θy4, the center position G and inclination of the component 9 are recognized by calculation.

FIG. 8 shows the recognition field of view 7 at each moving position at the same time, and L-α indicates the amount of movement of the recognition field of view 7, and this movement allows the corner position and direction of each side to be determined with respect to the overall image of the part 9. is imaged, and from the relationship between these images and each recognized visual field position, it can be seen that the corner position and orientation regarding the entire image of the component 9 are uniquely determined. And from this decision part 9
It can also be seen that the true center position G in the entire image can be calculated.

As described above, even if the part 9 is too large to fit into the recognition field of view 7, by moving the recognition field of view 7 and sequentially taking images of each corner, the position and inclination of the part 9 can be determined as a whole image of the part 9. can be recognized in relation to

Note that in the recognition field of view 7, it is only necessary to know the corner and the inclination of the two sides forming this corner, so it is not necessarily necessary to image the entire part 9 by moving the recognition field of view 7. Therefore, if the recognition method described above is used, there is no limit to the size of the part to be recognized. The orientation of the parts may be recognized as necessary.

Further, in the above embodiment, the case of an electronic component having a rectangular pattern has been described, but the position and inclination of various shapes such as a triangular pattern can be similarly recognized. In this case, recognition of some of all corners can be omitted depending on the shape of the pattern.

[0023]

According to the present invention, a plurality of corners of a rectangular pattern are imaged in recognition fields of view set at different positions. The position information and the position information of each recognition field of view that recognizes the position information are elements that uniquely determine the corner position in the overall image of the rectangular pattern, so it is possible to respond to the overall image of the rectangular pattern from these information. The required position can be calculated and recognized with high precision, and is effective when the pattern is too large to fit within the recognition field of view.

[0024]Furthermore, since the imaging of the part only needs to be done at a necessary corner of the part and does not need to cover the entire part, the required position can be recognized without any restriction on the size of the pattern.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing a recognition device in which the present invention is implemented.

FIG. 2 is a flowchart showing the operating procedure of the recognition method of the present invention.

FIG. 3 is a flowchart showing the operating procedure of the recognition method of the present invention.

FIG. 4 is a recognition visual field diagram showing an imaging state of a first corner of a pattern.

FIG. 5 is a recognition visual field diagram showing an imaging state of a second corner of a pattern.

FIG. 6 is a recognition visual field diagram showing the imaging state of the third corner of the pattern.

FIG. 7 is a recognition visual field diagram showing the imaging state of the fourth corner of the pattern.

FIG. 8 is an explanatory diagram in which captured images in each recognition field of view in FIGS. 4 to 7 are superimposed.

[Explanation of symbols]

1 Video camera 2 Microcomputer 3 Drive device 4 Image recognition device 5 Program 6 Input device 9 Parts 11 First corner 12 Second corner 13 Third corner 14 4th corner 21-24 sides G Center position x1, y1~x4, y4 position

Claims (1)

[Claims] 1. A rectangular pattern image captured in each recognition visual field by setting a recognition visual field corresponding to a plurality of corners of the rectangular pattern with one corner of the rectangular pattern to be recognized as a unit imaging target. A multi-view recognition method for a rectangular pattern, characterized in that the required position of the rectangular pattern is calculated and recognized from corner position information on the recognition field of view and relative position information of each recognition field of view that recognizes the position information. .