JPH01155202A - Part mounting apparatus - Google Patents

Abstract

PURPOSE:To perform processing within a short time, by taking the image of the shape of a part to output an image signal and providing a Hough converter for applying Hough conversion to said signal within a retrieval angle range sufficient to discriminate the shape of the part. CONSTITUTION:The image of a lead wire is taken to obtain a signal through a binary converter and said signal is subjected to Hough conversion to detect the intersecting point on lead center lines crossing each other at a right angle by an intersecting point detector. With respect to (a) : the direction crossing the lead center lines of a lead group at a right angle, (b) : the directions of leads crossing each other at a right angle and (c) : the intersecting point calculated when there is no bending of the first lead, the intersecting point of the led subjected to a bending test and the lead crossing said lead at a right angle is transferred to the first testing point satisfying the conditions (a) - (c) of the distance, when said intersecting point does not pass the leading ends of the leads of (b) and is moved to the position near to said leading ends, to detect a pixel by the first inspection means and the first testing point is successively moved by the lead interval in the lead and part direction to inspect that a testing point where the pixel of a pat is not detected is continuously generated two times by the second inspection means. Further, an operator constituted, so that 3 is subtracted from the total inspection number of times of the first and second inspection means to detect the number of bent leads, is mounted.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a component mounting device for automatically mounting components.

[Conventional technology]

In recent years, automatic component mounting devices have become widespread and are used in various places. For example, this involves taking an image of a part, converting the image into binarization, recognizing the straight line, and determining a reference point based on that straight line.
K to find the installation coordinates of the parts based on that reference point.
It has become. At this time, in order to recognize the straight line portion from the binarized signal, the least squares method is used to reduce the error. [Problems to be solved by the invention] However, such conventional devices If there is noise in the signal, the contrast with the background is low, or there are depressions or bumps in the area where you are trying to detect a straight line,
This part became a source of error, and sometimes incorrect results could not be obtained.

[Means for solving problems]

In order to solve these problems, the present invention applies Hough transform with a limited search angle range.

[Effect]

Detection is performed that is resistant to noise.

〔Example〕

When positioning an electronic component such as an IC using a visual device, two reference points are set on the component in consideration of the rotation of the component. At this time, for example, in the case of an IC with leads coming out from four sides, two intersections between the leads on one side and the leads on a different side may be found and used as reference points.

FIG. 1 is a flowchart of the means for finding this reference point. In the figure, an image is first input as shown in step 10G. At this time, in order to ensure accuracy, only a part of the corner of Ic1 is imaged by the imaging device 2, as shown in FIG. The captured video is subjected to horizontal edge detection to detect vertical lines as shown in step 101, and subjected to Hough transformation as shown in step 102. From the result, calculate the angle θV as shown in step 103,
In step 104, the coordinates ρv1 to ρVm are determined. These ρ and θ are 1 as shown in FIG. 3. ρ is the length of the ninth power line drawn from the origin to the straight line α, and θ is the angle between the fourth power line and the X axis.

The straight line α in FIG. 3 can be expressed as follows.

ρ" corresponds to a straight line in the X-y space, and the locus in the ρ-θ space expressed by the inverse K
This represents all groups of straight lines passing through O*y').

A similar operation is performed on all edge points in the image to find the location (ρ0θ0) where the trajectories are concentrated in the ρ-θ space, as shown in FIG. This is X-y
This corresponds to the straight line ρ0=1(2)θo+y θolc in space, which means that a straight line has been detected.

From the results obtained in this way, as shown in step 105, a set of straight lines on both sides of the terminal is found based on data on the terminal width and the pitch between the terminals, and the center line of each lead is determined. Similar operations are performed for steps 111 to 115, and based on these data, as shown in step 116, the intersection of the center lines of the k-th terminals (Xp, Y
Perform the operation to obtain p). The intersection point at this time is shown by the following formula +21, +3) when each specification is defined as shown in FIG.

In this way, to find the intersection point, the leads that are orthogonal to each other (the leads at the ends are more likely to be bent than the other leads, so usually the second lead from the end or later are used) are displayed on the screen. It is necessary to be photographed. At this time, the parts are not positioned accurately, so in order for the leads that are orthogonal to each other to be sure to enter the screen, the parts must be photographed with some margin as shown in Figure 6. There is a need to. For this reason, if the screen size is 256 x 240 pixels, which is the mainstream for industrial use,
The lead length L is about 30 to 40 pixels.

In the Hough transform described above, the lead shown in FIG. 7(a) is imaged to obtain an edge image as shown in FIG. 7(b), and this image is binarized to generate a signal as shown in FIG. 7(e). Then, by performing a Hough transform, two straight lines a and b as shown in FIG. 7(d) are obtained, and this center line C is obtained as the center line of the lead. This method using the Hough transform has the advantage of being resistant to noise compared to other methods, such as the least squares method, and is an effective method for detecting straight lines, but it takes a long processing time. Due to these disadvantages, it is currently not used in the industrial field.

However, when considering the positioning of electronic components, the component position can be determined mechanically to some extent, so the detection angle range when detecting a straight line (by Hough transform) is ±10
The amount of calculation can be reduced, and the processing time can be significantly reduced.

When positioning electronic components, it is common to set the one-view device so that the leads of the components are parallel or perpendicular to the vertical axis (y-axis) of the TV right camera, as shown in FIG. When detecting the center line of a lead in such an arrangement, it is necessary to perform a Hough transform within a range of ±10 with respect to the vertical and horizontal axes.As an example, if the center line of the lead is detected, a slope of 5.7 with respect to the vertical axis is required. Figure 8 (a) shows an analog image of a straight line with
) The K digital image is shown in Figure 8(b), but as mentioned above, the length of the lead can only be represented by about 30 pixels, so the digital image with a tilt of 5.7 degrees has a length of 1.0 pixels and a tilt of 1.0 pixels. 〇′ is represented by a straight line consisting of three lines at 0 degrees.Therefore, by Hough-transforming this, in addition to the detection angle θ=5.7 that should be found, a peak also appears at 1θ=θ°, and θ=0° and θ
Two straight lines with an angle of =5.7° are detected. Furthermore, for straight lines with small inclination angles of 1° and 2°,
If the influence of peak #20 is even stronger, the linear peak that should be detected is omitted, resulting in an undetectable result.

As described above, when a straight line to be detected is short, such as when detecting a lead of an electronic component, a problem arises when attempting to detect it using the Hough transform. This means that the slope of the straight line to be detected is
Not only when the right camera is near 0 degrees with respect to the vertical (horizontal) axis, but also in the range of 0 to 90 degrees, the pixels are arranged vertically or horizontally as shown in Figures 9 (a) to (1). When they line up, K also occurs. In Figure 9, -) is 0 degrees, (b) is 90 degrees,
(C) is 45 degrees, (d) is 26.6 degrees, (・) is 63 degrees.
4 degrees, (f) is 18.4 degrees, ■ is 71.6 degrees, (h) is 33.7 degrees, and (1) is 56.3 degrees.
1) to (0) are angles with large influence, (a), (@
) is an angle with a medium degree of influence, and (f) to (1) are angles with a small degree of influence.

When electronic components are mounted, the angle to be detected is limited to a certain range (usually about ±10 degrees), so as shown in Figure 10, the TV right camera is perpendicular to the lead of the component. If the (horizontal) axis is tilted by 14 degrees (or 76 degrees), as shown in FIG. 9, it becomes possible to detect straight lines within a range excluding angles without adversely affecting straight line detection. This is illustrated in FIG. 11.

In the above examples, the tilting angle is 14 degrees or 76 degrees, but this is within the search angle range (±10 degrees in the example in Figure 11).
0 degrees, 90 degrees, 45 degrees, and 26.6 degrees, which have a large influence on
It is sufficient to set the angle so that the angle does not exceed 63.4 degrees.

By the way, when detecting the center of an IC lead as shown in Figure 5, the length of the lead wire is short, so in order to improve the detection angle accuracy, all the lead wires in the same direction are assumed to have the same slope, and the average Detecting the angle. By processing in this way, the same effect as detecting the angle of a long object can be obtained. However, when some of the multiple leads have different slopes and deviate greatly from the average, they are considered to belong to a different group and are excluded from the detection results.

For example, as shown in FIG. 12, when there is a set of leads indicated by the symbol "A" and a set of leads indicated by the symbol "mouth", the intersection of the leads should normally be at the triangular mark in the figure. However, in the Hough transform, the angle of the lead inclination is detected as an average, so the first lead in the group marked with the symbol "A" has a larger inclination than the average, and is recognized as belonging to a different group. The intersection point is recognized as being at the black circle mark. Once this kind of recognition is made, it is difficult to correct this mistake in the future, so it is necessary to detect that the lead is significantly bent before that happens.The following method is effective. Become.

To explain with reference to FIG. 12, first, when examining the bending of the leads of the group marked with symbol "A", the intersection point P1 between the first bins is
(Originally, this point is not the intersection of the 1st bin and the 1st bin, but since the 1st bin was crooked and was not recognized, the result for this inspection was that the 1st bin was the 1st bin.) Using this as a reference, a point P2 is set that is moved in a direction perpendicular to the lead line for which this intersection Pl is to be detected. At this time, the moving distance between points P1 and P2 is the original first IJ
- When moving the wire at the intersection (the leftmost intersection indicated by a triangle mark) of the leads (bent leads in the figure), it does not pass the tip of the "mouth" of the lead pair, and the tip of the lead wire Make the points as close as possible. And that point P
Assume a line Ll drawn parallel to the direction of the lead wire group "A" from 8.0 As mentioned above, after each lead is imaged, it is binarized and the pixel of a certain part of the lead is is white, and pixels in areas without leads are black. Point P
2 is on the horizontal lead center line of the first □, so
The pixel at that position is "white". Then, if we check the number of white pixels in the direction from point P2 to line L1, since point P2 is on the lead, there should be a certain number or more of white pixels.

Next, the point P2 is moved outward by the distance l between the lead wire centers and the same inspection as the previous time is performed. At this time as well, if the first lead of the group marked with symbol "A" in Figure 12 is not bent, point P3 will jump out from the horizontal lead line, so the point P3 at that time will be
Pixel number 3 should be determined to be black. However, the 12th
In the example shown in the figure, since there is a bend in the lead, point P3 exists on the lead in the horizontal direction, and the pixels there are determined to be white.

Next, if point P3 is moved to the left by the distance between the lead lines and set as point P4, point P4 will be located off the top of the first lead in the horizontal direction. Therefore, the pixel at that point is determined to be black. Then, point P4 is again moved to the left by the lead line interval and the same detection is performed to confirm that the pixel at that point is also black. I have performed detection four times so far.
The last two times are to confirm that black pixels appear twice in a row. Therefore, the inspection ends when it is detected that the determination result of no lead wire is present twice in a row, and the number of lead wires that could not be detected is the number obtained by subtracting 3 from the number of inspections up to that point. In the example shown in FIG. 12, assuming that the lead songs of the first and second pins of the set with the symbol "A" have a large A, the number of tests will be 5 if the above-mentioned test is performed. Therefore, the number of bent lead wires is 2, which is obtained by subtracting 3 from the number of inspections (5).

This test is carried out one after another, and when it is completed, the symbol "mouth" is displayed.
If the test for the set of 2 is carried out in the same way, bends in all the leads can be detected.

〔Effect of the invention〕

As explained above, since this invention performs the Hough transform by limiting the search angle range, it has the effect of realizing processing in a short processing time while taking advantage of the noise resistance peculiar to the Hough transform. have

[Brief explanation of the drawing]

Fig. 1 is a 70-chart showing an embodiment of the present invention, Fig. 2 is a perspective view showing an imaging situation, Fig. 3 is a diagram showing the relationship between ρ and θ, and Fig. 4 is a ρ-θ space. Graph showing the relationship between
Figure 6 is a diagram explaining the intersection of leads, Figure 6 is a diagram explaining the imaging range, Figure 7 is a diagram explaining Hough transform, Figure 8 is a diagram showing the imaging result of slope lines, and Figure 9 is a diagram explaining the influence. Figure 10 is a diagram explaining the angle of inclination of the camera. Figure 11 is a diagram showing the relationship between the angle of inclination of the camera and the degree of influence.
FIG. 2 is a diagram showing a method for inspecting lead bending. 1...IC12...imaging device. Patent applicant: Yamaha Motor Co., Ltd. Agent: Masaki Yamakawa (2 employees) Figure 1 Figure 2 Figure 5 Figure 6 Figure 7 (Q) (b)

Claims (3)

[Claims] (1) A component mounting device that clamps a component and attaches it to a predetermined position includes an imaging device that images the shape of the component and outputs an image signal, and a device that uses the obtained image signal to identify the shape of the component. A component mounting device comprising: a Hough transformer that performs a Hough transform in a sufficient search angle range. (2) A component mounting device that clamps a component and mounts it in a predetermined position includes an imaging device that images the shape of the component and outputs an image signal, and an x-axis or y-axis of the imaging device that is connected to the x-axis of the component. Alternatively, it is composed of a rotation device that rotates within a predetermined range with respect to the y-axis, and a Hough transformer that performs Hough transform on the obtained image signal within a search angle range sufficient to identify the shape of the part. The rotation angle is 0 degrees, 26.6 degrees, 45 degrees, 63.4 degrees in the search angle range with respect to the x-axis or y-axis of the imaging device.
A component mounting device characterized in that the value does not include degrees or 90 degrees. (3) After confirming that the leads are planted on each perpendicular surface, that the leads on each surface are lined up in parallel, and that the bending is within a specified range, the part is placed in a predetermined position. The component mounting device that is attached to the device includes an imaging device that images the lead wire and outputs an image signal, a binarization converter that binarizes the image signal, and a Hough transform of the obtained binarized signal. An intersection detector that detects intersections on orthogonal lead center lines, and an intersection between a lead to be tested for bending and a lead orthogonal to it (a) in a direction perpendicular to the lead center line of a group of leads including the lead to be tested (b) The direction in which the lead is perpendicular to the lead to be inspected (c) The distance required when moving the point of intersection obtained when the first lead is not bent, without passing the tip of the lead in (b) and moving to a position close to the tip. The first test point is moved to the first test point that satisfies all conditions (a) to (c) of
and a second inspection means for sequentially moving the first test point toward the lead end by the lead interval and inspecting whether a test point where no component pixel of the component is detected occurs twice in a row. A component mounting device comprising: a computing unit that detects the number of bent leads by subtracting 3 from the total number of inspections by the first and second inspection means.