JPH0755429A - Position recognition system - Google Patents

Abstract

PURPOSE:To allow highly accurate position detection even if the fiducial mark of a substrate has a chip or the size thereof fluctuates. CONSTITUTION:An image processor for processing input images from an ITV camera picking up the image at the fiducial mark part of a substrate has an image data creating function for digitizing the input image, a template image memory function for storing the data of template image T, and a function for detecting the position of fiducial mark through matching. The template image T comprises a circular black region B corresponding to the central fiducial mark, a gray region G having gray level decreasing gradually toward the outside, and the outermost white region W. The degree of correlation increases as the center of template T comes near the center of fiducial mark.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is incorporated in, for example, a component mounting apparatus for mounting components on a board, and recognizes the position of the fiducial mark on the basis of photographing the fiducial mark on the board. Regarding the recognition device.

[0002]

2. Description of the Related Art For example, in a component mounting apparatus or the like, the position of a positioning fiducial mark provided on the surface of the substrate is detected by a position recognizing device so that the position of the substrate is adjusted. Has become. In FIG.
Some examples of this type of physical mark F (shaded for convenience) are shown, generally circular,
Various patterns such as triangles and squares are used.

The position recognition device captures the physical mark F by an ITV camera, converts the input image I into binary data, and detects the position coordinates of the physical mark F by the template matching method. It has become. In this template matching method, as shown in FIG.
A template image T having a pattern equivalent to that of the mark F is prepared for each type of the mark F, and the position of the template image T having the highest degree of correlation between the input image I and the template image T is calculated to calculate the fiducial mark F. The position of is detected.

[0004]

The physical mark F is formed at the same time when the conductor pattern of the substrate is formed.
As shown in 0 (a), a part of the outer edge part is chipped,
As shown in FIGS. 10B and 10C, the size may vary.

In the template matching method using the template image T having the same pattern as that of the conventional physical mark F as described above, if the physical mark F has such a chipping or size variation, There is a problem that the accuracy of detecting the position of the charmark F is reduced. In particular, when the size of the physical mark F and the pattern of the template image T do not match, a plurality of positions having the highest degree of correlation appear as shown in FIG. Since the position deviated from the position is detected, the reliability of the position detection becomes low.

The present invention has been made in view of the above circumstances, and its object is to recognize the position of a physical mark on a substrate, and when the physical mark has a chip or a variation in size. However, it is an object of the present invention to provide a position recognition device capable of highly accurate position detection.

[0007]

The position recognition device of the present invention comprises a photographing means for photographing a fiducial mark, an image data generating means for digitizing an input image by the photographing means, and the fiducial mark. A template image storage unit that stores a template image of a corresponding pattern, and a position detection unit that performs a correlation calculation between the input image and the template image and detects the position of the physical mark based on the degree of correlation. Then
The template image is characterized in that density gradations that gradually change from the inner side to the outer side are expressed corresponding to the edge portions of the fiducial mark.

In this case, the template image having a circular pattern can be commonly used when detecting the positions of the physical marks having a pattern other than the circular pattern. Further, the template image may be a binarized image, and the density gradation of the region may be equivalently expressed by the density of the binary pixel pattern in the region consisting of several pixels. Furthermore, it is more effective to provide a self-help generation function for generating a template image from a plurality of input images of fiducial marks.

[0009]

According to the above means, the template image is gradually processed from the inner side to the outer side in correspondence with the edge portion of the fiducial mark, that is, the portion where the fissure mark may be chipped or the size may vary. Since the density gradation that changes is expressed, by weighting the correlation calculation according to the density of the template image, the closer the center of the template image pattern approaches the center of the physical mark of the input image, the closer The degree of correlation becomes high. As a result, it is possible to detect the fiducial mark with high accuracy even if the fiducial mark is chipped or the size varies.

In this case, if the template image has a circular pattern, the center of the template image pattern is located at the center of the fiducial mark of the input image, even if the fiducial mark has a pattern other than the circular shape. The closer it gets, the higher the correlation becomes, so it can be commonly used to detect the position of a non-circular physical mark, and a template image for each pattern of the physical mark is prepared. There is no need to do it.

Further, if the template image is a binarized image, and the density gradation of the region is equivalently expressed by the density of the binary pixel pattern in the region consisting of several pixels, the fidelity is similarly obtained. It is possible to detect the juicy mark with high accuracy, and at the same time, it is possible to shorten the processing time and simplify the structure as compared with the case of using the multi-valued data.

Furthermore, by providing a self-assisted generation function for generating a template image from a plurality of input images of fiducial marks, it is not necessary to previously create a template image expressing density gradation from theoretical data. It is possible to automatically generate a template image that expresses density gradation according to the occurrence of actual chipping of the physical mark and the variation in size.

[0013]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A first embodiment of the present invention will be described below with reference to FIG.
It will be described with reference to FIGS. FIG. 1 shows a schematic configuration of a position recognition device 1 according to this embodiment. The position recognition device 1 is incorporated in, for example, a component mounting device that automatically mounts components on the substrate 2, and is roughly classified into a light source 3 for illuminating the substrate 2 and a position alignment formed on the substrate 2. The ITV camera 4 is a photographing means for photographing the portion of the fiducial mark F for use in, and an image processing device 5 for processing an input image I (see FIG. 8) from the ITV camera 4. In this embodiment, the physical mark F is circular, for example, as shown in FIG.

The image processing device 5 includes an A / D (not shown).
It is composed of a converter, a microcomputer, etc., and has each function as shown in FIG. That is, the image data generation function is for digitizing the input image I from the ITV camera 4 and converting it into binary data in this case. For example, the density of each pixel of the input image I of M × M pixels, Binarization is performed with a black level and the other part (ground part of the substrate 2) as a white level. The binarized data of the input image I is stored in the image memory.

The template image storing function stores the data of the template image T necessary for detecting the physical mark F by the template matching method. As shown in FIG. 2, the template image T has a pattern corresponding to the physical mark F in a frame of N × N pixels (N <M), for example. Has a circular black region B (shown with hatching for convenience), and a gray region G in which the density gradually becomes lighter from the inside to the outside in the peripheral portion thereof.
(Illustrated by a plurality of concentric circles for convenience), and has a white region W at the outermost peripheral portion.

In this case, the size of the black area B is set smaller than, for example, the smallest physical mark F to be photographed, and the size of the outer peripheral portion (the circle located at the outermost circumference) of the gray area G is For example, it is set larger than the maximum fiducial mark F to be photographed. Therefore, the gray area G is provided corresponding to the edge portion of the fiducial mark F to be photographed. Also,
The density value of each pixel is represented by 16 gradations from 0 (white level) to 15 (black level), for example.

The position detecting function performs a correlation operation between the input image I and the template image T by, for example, the residual sequential verification method (SSDA method), and determines the position of the financial mark F based on the degree of correlation. It is designed to detect. Therefore, the image processing device 5 functions as the image data generating means, the template image storing means, and the position detecting means according to the present invention. Further, in this embodiment, as will be described later in detail, the image processing device 5 is
The template image T is automatically generated from a plurality of input images I of the fiducial mark F by a self-help generation function.

Now, the flow chart of FIG. 3 shows a procedure for detecting the physical mark F on the substrate 2 in the position recognition apparatus 1 configured as described above. First,
In step S1, the ITV camera 4 photographs the part of the fiducial mark F on the substrate 2 that has been carried into the predetermined position, and in step S2, the input image I is digitized. The input image I is as shown in FIG. 8A, and the mark F portion (shown by hatching for convenience)
Pixels are converted into a black level and the other pixels are converted into a white level into binary data.

In step S3, the correlation calculation between the input image I and the template image T is executed. In this calculation, when the template image T is superposed on the input image I and the density values of the corresponding pixels are Ti and Ii respectively, the Sp value of Sp = Σ | Ti-Ii | It is performed by moving the input image I and moving the input image I. At this time,
As described above, the density value Ti of each pixel of the template image T takes a 16-step numerical value of 0 to 15, and the density value Ii of each pixel of the input image I has either a white level of 0 or a black level of 15. Takes the value of. In this case, also for the input image I, the density of each pixel may be represented by multivalued data.

When the value of Sp at each position of the template image T is obtained, the S value is calculated in step S4.
The position (center position) of the physical mark F is obtained based on the position of the template image T where the p value is the smallest (the degree of correlation is the largest). Incidentally, if Sp exceeds a predetermined threshold value during the addition when obtaining Sp, the calculation is terminated at that point.

In this case, the black area B of the template image T
The black level portion (fiducial mark F) of the input image I overlaps the whole, and the white level portion of the input image I (the base portion of the substrate 2) covers the entire white area W of the template image T.
And the black level portion (fiducial mark F) of the input image I overlaps the inner portion of the gray area G of the template image T as much as possible, the correlation degree becomes maximum.

Therefore, the closer the center of the pattern of the template image T is to the center of the physical mark F of the input image I, the higher the degree of correlation becomes. As a result, the physical mark F is missing or large. Even if there is a variation in height (see FIG. 10), the detection of the fiducial mark F is performed with high accuracy.

It should be noted that the residual sequential verification method (SSDA) is used here.
However, other methods such as the method using the cross-correlation coefficient and the MF method (generalized Houg) are used.
h conversion), thinning method, pyramid structuring method, and other various methods can be adopted.

Next, the self-help generation function will be described with reference to FIG. The self-help generation function is a function of extracting and creating the template image T from a plurality of actual input images I obtained by photographing the portion of the fiducial mark F, and FIG. 4 shows the procedure thereof.

First, the positions of a plurality of input images I are aligned (P1). This is a process for eliminating the misalignment of the physical mark F portion of each input image I in the X, Y and θ (rotation) directions. Next, the density values of the respective pixels in the state where the registered input images I are superimposed are added (P2). With this, the appearance probability of each pixel is obtained. The density value of each pixel before addition may be binary data or multivalued data.

Then, the density values of the added pixels are normalized (P3). This normalization is performed by setting the maximum value of the added density values to 100% and the minimum value to 0%, and setting the density value of 100% to 0% to each pixel. Finally, the image is cut out in a frame of N × N pixels (P4) to obtain the template image T shown in FIG. According to this, it is possible to obtain the template image T having the appropriate gray area G in consideration of the chipping of the physical mark F that actually appears and the variation in size.

As described above, according to the present embodiment, the template image T is made to express the density gradation which gradually changes corresponding to the edge portion of the physical mark F. Even if there is a situation where a part of the outer edge part is chipped or the size varies, unlike the conventional one, highly accurate position detection can be reliably performed, and the reliability is greatly improved. be able to.

Moreover, particularly in the present embodiment, since the self-help generating function for generating the template image T from the plurality of input images I of the fiducial mark F is provided, the template image is not created from theoretical data in advance. In addition, it is possible to automatically generate the template image T that expresses the density gradation according to the occurrence of chipping of the actual physical mark F and the variation in size.

FIGS. 5 (a) and 5 (b) respectively show template images T according to the second embodiment of the present invention. This embodiment is different from the first embodiment in that the template image T is composed of a binarized image, and the density gradation is equivalently expressed by the density of the binary pixel pattern.
In this case, in order to create the template image T, in the step P3 of the self-help generation function in the first embodiment,
Assuming that the density value of 100% is the black level (1) and the density value of 0% is the white level (0), for example, in the area of the density value of 50%, 0 and 1 are repeated every two pixels, and the area of the density value of 25%. Can be obtained by creating a dither image (coarse / fine image) in which 3 pixels are set to 0 for every 4 pixels and 1 pixel is set to 1.

According to this structure, the same operation and effect as those of the first embodiment can be obtained, and the advantages such as the improvement of the operation processing speed and the saving of the image memory can be obtained. You can

FIG. 6 shows a template image T according to the third embodiment of the present invention, in which the density gradation changes stepwise from the inner side to the outer side. For example, a small circular area at the center is a black area B having a density value of 100%, an outer ring area is a gray area G1 having a density value of 70%, and an outer ring area is a gray area having a density value of 30%. G2 and the outermost area are white areas W having a density value of 0%. Even with such a configuration, the same operation and effect as those of the first embodiment can be obtained.

Although not described in detail, the template image T having a circular pattern as shown in FIGS. 2, 5 and 6 is, for example, a triangle shown in FIG.
For example, the (c) quadrangle or the (d) quadrangle rotated by 45 degrees,
Even a regular polygonal fiducial mark F other than a circle can be commonly used.

That is, since the center of the inscribed circle and the center of the circumscribed circle coincide with each other in the triangular and quadrangular fiducial marks F, the ring-shaped portion between the inscribed circle and the circumscribed circle is included. If the gray area G is included, the position can be detected by the same processing as the method of FIG.
Even in this case, the fiducial mark F of the input image I
The closer the center of the pattern of the template image T is to the center of the image, the higher the degree of correlation is, and the detection of the physical mark F is performed with high accuracy even if the physical mark F has a chip or a variation in size. It is possible.

Finally, FIG. 7 shows a fourth embodiment of the present invention. When detecting the position of the ring-shaped (hollow circle) fiducial mark F as shown in FIG. 7A, the input image I is binarized as shown in FIG. 7B. An inverted image J that has been inverted is generated, and template matching is performed on the circular black level portion at the center of the inverted image J. At this time, although not shown, the template image expresses density gradations that gradually become lighter from the inner side to the outer side in correspondence with the edge portion of the circular black level portion in the central portion. As a result, the position of the physical mark F can be detected with high accuracy as in the first embodiment.

When the position of such a hollowed-out form of the fiducial mark F (see FIGS. 8 (e) to 8 (h)) is to be detected, the template image is set to the outer edge portion of the fiducial mark F. The density gradation is expressed such that it gradually becomes lighter from the inner side to the outer side in correspondence with the portion, and the density level becomes gradually lighter from the outer side to the inner side in correspondence with the inner edge portion of the fiducial mark F. It may be configured to express a key.

Besides, the present invention is not limited to the above-mentioned respective embodiments, and the template image T may be theoretically created without depending on the self-assisted generation function,
Further, the present invention can be implemented by appropriately modifying it within a range not departing from the gist such that it can be incorporated in various devices other than the component mounting device.

[0037]

As is apparent from the above description, according to the position recognition device of the present invention, the position of the physical mark on the substrate is recognized based on the correlation calculation between the input image and the template image. Yes, in the template image,
The density gradation that gradually changes from the inner side to the outer side is expressed corresponding to the edge portion of the physical mark, so that high accuracy is achieved even when the physical mark has a chip or a size variation. This has an excellent effect that the position detection can be performed.

In this case, a template image having a circular pattern can be commonly used for detecting the position of a non-circular physical mark, and it is necessary to prepare a template image for each pattern of the physical mark. It is possible to obtain an advantage such as disappearance.

Further, if the template image is a binarized image and the density gradation of the region is equivalently expressed by the density of the binary pixel pattern in the region consisting of several pixels, it is multi-valued. Processing time can be shortened and the structure can be simplified as compared with the case of using data.

Furthermore, by providing a self-assisted generation function for generating a template image from a plurality of input images of fiducial marks, it is not necessary to create a template image expressing density gradation from theoretical data in advance. It is possible to automatically generate a template image that expresses density gradation according to the occurrence of actual chipping of the physical mark and the variation in size.

[Brief description of drawings]

FIG. 1 shows a first embodiment of the present invention, and is a diagram schematically showing the overall configuration.

FIG. 2 is a diagram showing an example of a template image.

FIG. 3 is a diagram showing a procedure for detecting a fiducial mark.

FIG. 4 is a diagram showing a processing procedure of a self-help generation function.

FIG. 5 is a diagram showing a template image in the second embodiment of the present invention.

FIG. 6 is a diagram showing a template image according to a third embodiment of the present invention.

FIG. 7 is a diagram for explaining a matching method according to a fourth embodiment of the present invention.

FIG. 8 is a diagram showing some examples of fiducial marks.

FIG. 9 is a diagram showing a template image in a conventional example.

FIG. 10 is a diagram showing the appearance of missing (a) and size variations (b) and (c) of the physical mark.

FIG. 11 is a diagram showing how a recognition result shift occurs due to variations in the size of a financial mark.

[Explanation of symbols]

In the drawing, 1 is a position recognition device, 2 is a substrate, 3 is a light source, 4 is an ITV camera (imaging means), 5 is an image processing device (image data generation means, template image storage means, position detection means), and I is an input. Image, T is a template image, F is a fjordule mark, B is a black area, W is a white area, and G, G
1 and G2 indicate gray areas.

Claims (4)

[Claims] 1. A position recognition device for recognizing the position of the fiducial mark on the basis of photographing a fiducial mark on a substrate, the photographing means photographing the fiducial mark, and the photographing means. Image data generation means for digitizing the input image, template image storage means for storing the template image of the pattern corresponding to the fiducial mark, and correlation degree between the input image and the template image Position detecting means for detecting the position of the fiducial mark on the basis of the above, wherein the template image has density levels gradually changing from the inside to the outside in correspondence with the edge portion of the fiducial mark. A position recognition device characterized by expressing a key. 2. The position recognition according to claim 1, wherein the template image having a circular pattern is commonly used even when the positions of the fiducial marks having a pattern other than the circular shape are detected. apparatus. 3. The template image is composed of a binarized image, and the density gradation of the region is equivalently expressed by the density of the binary pixel pattern in the region consisting of several pixels. The position recognition device according to item 1 or 2. 4. The position recognition device according to claim 1, further comprising a self-help generation function for generating a template image from a plurality of input images of the fiducial mark.