JP2003208616A - Image recognition device and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To compensate a rotation angle (θ3) around Z-axis with one camera. <P>SOLUTION: An image recognition device is provided with a camera 2 for photographing an object 1a to be recognized, a storage means 16a for storing a template for the object 1a to be recognized and position data of the object to be recognized in an image of the camera, and an image recognition means 13a for performing pattern matching by the image of the camera 2 and the template. The image recognition means 13a performs a pattern matching in two different regions on one image, calculates the position of a feature point in respective regions, and obtains an angle formed by a straight line passing two indexed feature points and a straight line determined by the position data and passing two feature points to use the obtained angle as a compensation amount of the rotation angle. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention
Not only the correction in the axial direction and the correction in the Y-axis direction, but also the correction in the Z-axis direction (correction of the distance) and the rotation angle around the Z-axis (here, θ
3)) and an image recognition device and a program capable of performing the correction.

[0002]

2. Description of the Related Art Conventionally, for automatic check (inspection) of a printed circuit board (circuit board) and mounting of parts, a robot hand is placed directly above a specific portion of the printed circuit board fixed on a horizontal stage. In an apparatus or the like that moves and extends a probe pin or the like for electrical measurement or mounting of components, it is necessary to accurately control the robot hand.

In many cases, such accurate positioning accuracy of the robot hand cannot be ensured, and some kind of correction is required. For this reason, the method that is often used is to perform some image recognition using the image captured by the camera attached to the robot's hand, and calculate how much the recognition result deviates from the original position. Then, the hand is adjusted (corrected) by an amount corresponding to it. If the surface of the stage on which the circuit board to be checked is fixed is the XY plane, this correction is composed of two corrections, the X-axis direction correction and the Y-axis direction correction.

Conventional image recognition requires a process for determining which part of a certain image corresponds to which part of another image. Such processing is performed by pattern matching (patter
n matching), and there is matching between images and matching between figures. It is also a problem of matching to find out where in the image a pattern corresponding to a specific object exists and recognize the object. In particular, the pattern of an object is expressed as an image, and the operation of determining the position of the object by examining the similarity between the pattern and each part of the object image is performed by template matching (template mat
ching), which is the most basic and commonly used method of matching in image processing.

FIG. 13 is an explanatory view of a conventional example. FIG. 13 is an explanation of the conventional template matching,
A template and image A are provided. In template matching, the center of the template representing the object to be detected is overlapped with the point (i, j) in the image A, the similarity between the template and the partial pattern of the overlapping image is measured, and the value is determined as the point. This is a method of determining the certainty that the target exists at (i, j). In order to obtain the position of the object, this operation is performed for all points (pixels) in the image, and the position where the degree of similarity is maximum is obtained.

[0006]

In the above-mentioned conventional device, the conventional correction is insufficient when the error in the Z-axis direction of the hand and the rotation angle (θ3) about the Z-axis cannot be ignored. For these two corrections, it is possible to install two cameras and make corrections using two images, but problems such as the need for two cameras and the large and heavy hands. was there.

It is an object of the present invention to perform correction in the Z-axis direction (distance correction) and correction of the rotation angle around the Z-axis (θ3) with one camera.

[0008]

FIG. 1 is an explanatory diagram of an image recognition apparatus of the present invention. In FIG. 1, 1a is an object to be recognized, 2 is a camera, 4 is a probe, 10 is a control unit, 11a is a drive unit, 13a is an image recognition unit, and 16a is a storage unit.

The present invention has the following means in order to solve the above conventional problems.

(1): A camera 2 for photographing the recognition object 1a, a storage means 16a for storing a template for the recognition object 1a and position data of the recognition object in the image of the camera, and the camera. Two images and an image recognition unit 13a that performs pattern matching with the template are provided. The image recognition unit 13a performs pattern matching in two different areas on one image, and the position of the feature point in each area. Is calculated, and the angle formed by the straight line passing through the two characteristic points thus calculated and the straight line passing through the two characteristic points determined by the position data is obtained, and the obtained angle is taken as the correction amount of the rotation angle. Therefore, the correction amount of the rotation angle can be easily obtained from one image.

(2): A camera 2 for photographing the recognition object 1a, a storage means 16a for storing a template for the recognition object 1a and position data of the recognition object in the image of the camera, and the camera. Two images and an image recognition unit 13a that performs pattern matching with the template are provided. The image recognition unit 13a performs pattern matching in two different areas on one image, and the position of the feature point in each area. Is calculated, and an error in the distance between the camera 2 and the recognition object 1a is calculated from the calculated distance between the two characteristic points and the distance between the two characteristic points determined from the position data, and the error is calculated as Use as a correction value. Therefore, the correction value of the distance can be easily obtained from one image.

(3): A camera 2 for photographing the recognition object 1a, a storage means 16a for storing a template for the recognition object 1a, and position data of the recognition object in the image of the camera, and the camera. Two images and an image recognition unit 13a that performs pattern matching with the template are provided. The image recognition unit 13a performs pattern matching in two different areas on one image, and the position of the feature point in each area. And the angle formed by the straight line passing through the two characteristic points thus determined and the straight line passing through the two characteristic points determined by the position data is determined, and the determined angle is used as the correction amount of the rotation angle, and is also determined. The distance between the two feature points and the distance between the two feature points determined by the position data are compared to measure the error in the distance between the camera 2 and the recognition object 1a. And, the correction value of the distance said error.
Therefore, the rotation angle correction amount and the distance correction value can be obtained from one image.

(4): In the image recognition device of (1) to (3), the image recognition means 13a corrects the correction amount of the rotation angle or the correction value of the distance, and then
The positions of the characteristic points after the correction and the positions of the characteristic points determined by the position data are corrected. For this reason, since the XY-axis correction is performed by eliminating the error of the rotation angle and the error of the distance in one image, it is possible to easily and accurately perform the correction of the hand including the camera.

[0014]

BEST MODE FOR CARRYING OUT THE INVENTION The present invention uses one camera and uses image recognition to perform not only correction in the X-axis direction and Y-axis direction but also correction in the Z-axis direction (distance correction) and Z The rotation angle (θ3) around the axis is also corrected.

(1): Description of Image Recognition Device FIG. 1 is an illustration of the image recognition device. In FIG. 1, the recognition target 1a is an IC lead or the like of a printed circuit board on which an electronic component is mounted. The camera 2 is attached to a robot hand or the like to obtain an image of the recognition object 1a.
The probe 4 is attached to a robot hand or the like to inspect the recognition target object 1a. The control unit 10 controls the entire image recognition apparatus. Drive means 11
a is an X-axis, Y-axis, Z-axis, Z that drives the robot hand
It is composed of four drive units for axial rotation (θ3) and the like.
If necessary, X-axis rotation (θ1), Y-axis rotation (θ2), etc. of the hand are provided, but they are not included here for simplicity. The image recognition means 13a performs processing such as pattern matching. The storage unit 16a stores CAD data (positional data) corresponding to the image of the camera 2, bright / dark (shading) data of the template, and the like. Less than,
An image recognition device applied to the automatic board checking device will be described.

(2) Description of automatic board checking: Description of automatic board checking device configuration FIG. 2 is an explanatory view of the automatic board checking device configuration. In FIG. 2, the printed circuit board 1, the camera 2, the hand (robot hand) 3, the probe 4, the probe pin 5, the stage 6, the motor x, the motor y, the motor z, and the motor θ3 are provided in the automatic board checking device. is there. The printed circuit board 1 is a check target on which electronic components are mounted. The camera 2 is attached to the hand 3 to obtain an image of the printed board 1. The hand 3 is driven by a motor to automatically check the printed circuit board.
The probe 4 is attached to the hand 3 to check the printed circuit board 1. The probe pin 5 makes contact with the contact point of the printed circuit board 1 for checking. The stage 6 is a base on which a check device is mounted. The motor x is a drive unit that moves the hand 3 in the X-axis direction. The motor y is a driving unit that moves the hand 3 in the Y axis direction (perpendicular to the paper surface). The motor z is a drive unit that moves the hand 3 in the Z-axis direction. Motor θ
Reference numeral 3 is a drive means for rotating the hand 3 in the Z-axis direction.

In FIG. 2, the number of the hands 3 is one for the sake of simplicity, but it is usually a plurality (generally about four).
The hand 3 is mounted, and at the same time, different points on the printed circuit board 1 are measured.

Description of Control Block of Automatic Board Checking Device FIG. 3 is an explanatory diagram of a control block of the automatic board checking device. The control block of the automatic board checking apparatus is provided with a control unit 10, a motor group drive unit 11, a camera control unit 12, an image recognition unit 13, a correction amount calculation unit 14, a measurement unit 15, and a measurement point data storage unit 16. .

The control unit 10 controls the entire system. The motor group drive unit 11 represents all the drive units of the motors required to move one hand. For example, X
It is composed of four motor drive units for axis, Y axis, Z axis, and Z axis rotation (θ3). The camera control unit 12 controls the camera 2 that captures an image of the printed circuit board 1. The image recognition unit 13 is an image recognition unit that performs image matching by performing pattern matching between a camera image and a template.
The correction amount calculation unit 14 calculates the correction amount of the hand from the image recognition result. The measuring unit 15 is, for example, a digital sampling oscilloscope, which measures voltage, electrical waveform, and the like. Measurement point data storage unit 16
Is a set of coordinate values of a position for contacting the measurement point of the printed circuit board (measurement object) 1 with the probe pin 5 or a set of data that can be converted into coordinate values (CAD data of the printed circuit board 1). ) Etc. are stored.

Description of Case in Which Z-Axis Rotation (θ3) is Required In a simple case, the θ3 motor is omitted and X, Y,
Contact at any point is possible with only three Z-axis motors. However, considering the following case, the motor of θ3 is required.

1) If there is an error around the θ3 axis when the printed circuit board 1 is attached, the X and Y axes of the printed circuit board 1 are aligned with the X and Y axes of the camera image by rotating the hand by θ3 (in this case, A motor of θ3 is not indispensable because rotation correction can be applied by image conversion.)

2) In order to avoid contact with an obstacle on the printed circuit board 1, the probe pin 5 is contacted in a tilted state. In this case, if the pretilted direction is the same as the direction in which the obstacle is present, it is possible to avoid the obstacle by rotating θ3.

FIG. 4 is an explanatory view of tilting the probe pin, FIG. 4 (a) is an explanation when the probe pin is not tilted, and FIG. 4 (b) is an explanation when the probe pin is tilted and rotated. In FIG. 4A, in the case where a large component 7 such as a hard disk is mounted on the printed circuit board 1, if the probe 4 is attempted to contact the point to be contacted without tilting, a part of the probe 4 interferes. Therefore, as shown in FIG. 4B, by rotating the hand 3 by θ3 with the probe pin 5 pre-inclined, the interference between the large component 7 and the probe 4 like the probes 4 on the left and right in the figure. Can be prevented.

3) When performing simultaneous measurement with a plurality of hands, if the contact points are close to each other, there is a risk of the hands colliding with each other. In this case, the collision may be avoided by rotating the hand.

4) When the hand has two pins, a probe pin and a ground pin. If the tip of the probe pin is aligned with the θ3 axis, both the probe pin and the ground pin can be properly contacted by properly moving the motor that moves the ground pin by a predetermined amount from the probe pin and the θ3 motor.

(3) Description of Indexing of Positions of Feature Points (XY Correction): Description of Using One Template FIG. 5 is an illustration of using one template. In FIG. 5, an image A and a template are provided. Image A is an image of the camera of printed circuit board 1,
An IC image a having leads around it is shown. The IC image a has leads (some leads are omitted) in the periphery,
These leads are connected to pads on the printed circuit board. The template is data (here, a part of the data of the read pattern) stored in advance for use in pattern matching.

(Description of Pattern Matching) In pattern matching, the approximate position of the image portion corresponding to the template (the image portion of the printed circuit board 1 fixed to the stage 6 in FIG. 2) corresponding to the template is known in advance from the CAD data. If there is no error, the pattern is supposed to match and the template is moved from that position to the front, back, left and right. FIG. 6 is an image recognition processing flow. Hereinafter, a description will be given according to the processes S1 to S3 of FIG.

S1: Set template data in the storage means of the image recognition unit 13 or the like. This data is data representing two-dimensional light and dark.

S2: The image recognition section 13 performs pattern matching between the template and the circuit board of the camera image to obtain the degree of similarity. This operation is repeated every time the template is moved by one pixel to create a two-dimensional similarity matrix. The range in which this template is moved is a range that covers the positional deviation error. The unit of movement is usually one pixel unit.

S3: The image recognition section 13 determines the pixel at the position of maximum similarity from the created similarity matrix and sets it as the matching position.

Description of using a plurality of templates FIG. 7 is an explanatory view of using two templates. In FIG. 7, an image A and a template are provided. The image A is an image of the camera, which has an IC image a having leads (some leads are omitted) and two VIs.
A (via) b is shown. Template,
This is data that is stored in advance for use in pattern matching (here, VIA and a part of the data of the read pattern). The VIA has a VIA hole (a hole (shown by a black circle) for connecting the layers of the multilayer printed circuit board) at the center. Further, although the template is shown here at the same time, the image recognition unit 13 is usually used.
It is stored separately in the storage means or the like.

(Explanation of Pattern Matching) In pattern matching, since the approximate position of the image portion corresponding to the template is known in advance from the CAD data, the position where the pattern is considered to be the same if the printed board has no positioning error is centered. Then, the template is moved back and forth, left and right from that position. FIG. 8 is an image recognition processing flow. Hereinafter, a description will be given according to the processes S11 to S15 of FIG.

S11: In the storage means of the image recognition unit 13,
Brightness data of the VIA template and the lead pattern template are set.

S12: The image recognition section 13 firstly detects the VIA.
And the circuit template of the camera image and the circuit board of the camera image (in the area centered on the position where the pattern will match if there is no error because the approximate position of the template is known in advance) and similar The similarity matrix is created by repeating the operation of obtaining the degree and the operation of moving the template by one pixel (pixel). The range in which this template is moved is set appropriately considering the positional shift range.

S13: Next, the image recognition unit 13 uses the template for the lead pattern and moves the pixel by one pixel in the same manner as the template to perform the pattern matching to obtain the similarity (the similarity matrix using the template create). This moving range is performed in the same range as the template (centering on the position where the patterns match if there is no error).

S14: The image recognition unit 13 next adds the similarity degree obtained using the template and the similarity degree obtained using the template at each position (the same movement distance) (the template and the template). The similarity matrix created using is added, and the added similarity matrix is created).

S15: The image recognition section 13 finds the pixel at the position of maximum similarity from the added similarity matrix and sets it as the matching position.

FIG. 9 is an explanatory diagram of similarity degree addition.
(A) is a description of the template similarity matrix,
FIG. 9B is a description of a template similarity matrix, and FIG. 9C is a description of a template and a similarity matrix obtained by adding templates.

The similarity matrix obtained by the template of FIG. 9A and the similarity matrix obtained by the template of FIG. 9B are added to obtain the result of FIG. 9C.
Is a similarity matrix of. For example, the similarities at the centers of FIGS. 9A and 9B are “18” and “20”, respectively.
Therefore, in FIG. 9C, the similarity at the center is “38”.

In FIG. 9C, "51" has the highest degree of similarity, and this position is one pixel right above the center (the position of CAD data). In this way, by using a plurality of templates, it is less likely to be affected by image noise and the like, and the recognition accuracy is improved.

If a template and a template having completely different patterns are used, it is expected that the recognition error will be further reduced (because there are cases where the error is likely to occur in different directions). Also, it is possible to use a large template such as a combination of templates, but the number of programs required for saving increases (the number of templates also increases),
This is not realistic because the processing amount of pattern matching increases.

In the above-mentioned embodiment, the description is made using the templates 1 to 2, but the use of three or more templates can further improve the recognition result of the image position.
Further, although the similarity is used for the pattern matching, the distance may be used. Contrary to the similarity, the distance becomes smaller (shorter) as the similarity becomes higher.

(4) Description of Z correction (distance correction) and θ3 correction using image recognition If only one camera corrects the X-axis direction (X correction) and the Y-axis direction (Y correction). Instead, the correction in the Z-axis direction (distance correction (Z correction)) and the rotation angle around the Z-axis (θ3) are also performed.

When the hand is moved to a predetermined position,
Determine the positions of the two feature points that would be visible on the camera. For this indexing, the indexing of the position of one point described above is used. For example, select one point each from the right half and the left half on the image of the camera. Here, the feature point refers to a point at which image recognition is possible. For example, it is a pad (foot pad) for connecting the lead of VIA or IC on the printed circuit board.

The original coordinates of the two feature points a and b are a = (x1, y1) and b = (x2, y2).

The coordinates of the two characteristic points A and B obtained as a result of image recognition by pattern matching are A = (X1, Y1) = (x1 + δx1, y1 + δy1) B = (X2, Y2) = (x2 + δx2, y2 + δy2) Hereinafter, a specific processing procedure of the Z correction and the θ3 correction performed by the image recognition unit 13 will be described.

Description of Rotation Angle (θ3) Correction The image recognition is performed separately in two different areas on one image, and the rotation angle correction amount is obtained from the result.

The rotation angle error δθ3 is calculated by using the vector cross product.

[0049]

[Equation 1]

FIG. 10 is an explanatory diagram of the correction of the rotation angle (θ3). In FIG. 10, the angle formed by the line segments a and b and the line segments A and B is δθ3. As a result, assuming that the error in the XY direction and the error in the height of the robot hand are "0", if the θ3 of the hand (the rotation angle around the optical axis of the camera) is rotated by δθ3 obtained here, the What was visible at points A and B turns in the opposite direction and is visible at points a and b, respectively. That is, θ3 is corrected by the rotation of δθ3 and becomes an accurate value.

Description of Height (Distance) Correction Image recognition is performed separately in two different areas on one image, and the correction amount of the distance to the recognition object is obtained from the result.

FIG. 11 is an explanatory diagram of height correction. Figure 1
1, the design height between the camera 2 and the object is h
Then, assuming that the actual height is h ′ and the error is δh = h−h ′, the hand (attached camera) can move to the correct height h by correcting the hand height by + δh. become.

[0053]

[Equation 2]

If the distance (design value) between the camera and the object is h, the actual distance is h ′, and the error is δh = h−h ′, then

[0055]

[Equation 3]

Description of Height (Distance) Correction and Rotation Angle Correction Image recognition is performed separately in two different areas on one image, and the results are used to determine the rotation angle correction amount and the recognition object. The amount of correction of the distance is obtained.

The correction of the rotation angle (θ3) and the correction of the height (distance) described above can be performed in combination.

[0058]

[Equation 4]

By correcting the error δθ3 of the rotation angle and the error δh of the height, both the rotation angle and the height can be made accurate.

Description of XY Correction and Rotation Angle Correction Image recognition is performed separately in two different areas on one image, and the XY correction amount and the rotation angle correction amount are obtained from the results.

(1) The rotation angle error δθ3 is determined by the method of correcting the rotation angle (θ3) described above (FIG. 12A).
reference).

(2) Next, the image of the camera is subjected to the rotation process of -δθ3. At this time, points A and B are points A'and B '
Suppose that it has moved to (see FIG. 12B).

[0063]

[Equation 5]

Instead of (2), the following process (2) 'may be performed.

(2) 'Rotate + δθ3 on the hand. After that, the image of the camera is acquired again, the image recognition is performed again, and the points of the recognition result are set as A ′ and B ′.

FIG. 12 is an explanatory diagram of XY correction and rotation angle correction, and FIG. 12A is an explanation of δθ3. 12
In (a), the angle formed by the line segment AB formed by the points A and B resulting from image recognition and the line segment ab formed by the originally (designed) points a and b is δθ3 (see (1) above).

FIG. 12B shows the description after the rotation processing.
In FIG. 12B, points A ′ and B ′ when the image is rotated by −δθ3 or the hand is rotated by + δθ3.
(See (2) or (2) 'above).

[0068]

[Equation 6]

Description of XY Correction and Height Correction The image recognition is performed separately in two different areas on one image, and the XY correction amount and the correction amount of height (distance) are obtained from the results. .

(1) The height error δh is obtained by the height correction method described above.

(2) Next, for the image of the camera, (h-
δh) / h times enlargement (or reduction) processing is performed. Here, the points after the conversion (processing) of the points A and B are A ′ and B ′.

[0072]

[Equation 7]

The following process (2) 'may be performed instead of the process (2).

(2) 'Move the hand by δh in the Z-axis direction. After the movement, the image of the camera is acquired again, the image recognition is performed again, and the points of the recognition result are set as A ′ and B ′. Then, XY correction is performed in the same manner as described above.

Description of Simultaneously Performing XY Correction, Rotational Angle Correction, and Height Correction Image recognition is performed separately in two different areas on one image, and from the results, XY correction amount, rotation angle correction amount, and high angle correction are performed. The amount of correction of the distance (distance) is obtained.

(1) The height error δh and the rotation angle error δθ3 are obtained by the above-described height correction and rotation angle correction methods.

(2) Next, with respect to the image of the camera, the rotation processing of −δθ3 described in (2) above and the above (2)
(H-δh) / h times enlargement (or reduction) described above
Perform processing. Here, the points after the conversion (processing) of the points A and B are A ′ and B ′.

[0078]

[Equation 8]

The following process (2) 'may be performed instead of the process (2).

(2) 'The hand is moved in the Z-axis direction by δh, and the rotation angle is also corrected by δθ3. After that, the image of the camera is acquired and the image recognition is redone, and the point of the recognition result is A ′,
B '. Then, XY correction is performed in the same manner as described above.

The method of obtaining the rotation angle correction, the height correction, and the XY correction can be performed as follows.

(1) The rotation angle, height, and X and Y correction amounts are calculated using the first camera image (these four calculation orders may be any order).

(2) After that, the XY correction amount is converted by the error of the rotation angle to obtain the final XY correction amount.

(3) Then, the rotation angle, height, and XY component (two components) are corrected for the mechanism (hand) (these four correction orders may be any order).

(5): Explanation of Program Installation Control unit 10, motor group drive unit 11, drive unit 11a, camera control unit 12, image recognition unit 13, image recognition unit 13
a, the correction amount calculation unit 14, the measurement unit 15, the measurement point data storage unit 16, the storage unit 16a, and the like can be configured by a program, are executed by the main control unit (CPU), and are stored in the main memory. It is a thing. This program is processed by a general computer. This computer is configured with hardware such as a main control unit, a main memory, a file device, a display device, and an input device that is an input means such as a keyboard.

The program of the present invention is installed in this computer. This installation is performed on a portable recording medium such as a floppy disk or magneto-optical disk.
These programs are stored and installed in a file device provided in the computer via a drive device for accessing a recording medium included in the computer or via a network such as a LAN. It Then, the program steps required for processing are read from the file device to the main memory and executed by the main control unit.

[Additional Notes] (Additional Note 1) A camera for photographing the recognition target object, a storage unit for storing the template for the recognition target object, and the position data of the recognition target object in the image of the camera. An image recognition unit that performs pattern matching with the image of the camera and the template, the image recognition unit performing pattern matching in two different regions on one image,
The position of the characteristic point is calculated in each area, the angle formed by the straight line passing through the two calculated characteristic points and the straight line passing through the two characteristic points determined from the position data is calculated, and the calculated angle is corrected for the rotation angle. An image recognition device characterized by a quantity.

(Supplementary Note 2) A camera for photographing an object to be recognized, a storage means for storing a template for the object to be recognized and position data of the object to be recognized in an image of the camera, an image of the camera and the An image recognition means for performing pattern matching with a template is provided, and the image recognition means performs pattern matching in two different areas on one image, and the position of the characteristic point is calculated in each area, and the calculated 2 An error in the distance between the camera and the recognition target is calculated from the distance between the two feature points and the distance between the two feature points determined by the position data, and the error is used as a correction value of the distance. Image recognition device.

(Supplementary Note 3) A camera for photographing an object to be recognized, a storage unit for storing a template for the object to be recognized and position data of the object to be recognized in an image of the camera, an image of the camera and the An image recognition means for performing pattern matching with a template is provided, and the image recognition means performs pattern matching in two different areas on one image, and the position of the characteristic point is calculated in each area, and the calculated 2 An angle formed by a straight line passing through one feature point and a straight line passing through the two feature points determined by the position data is determined, and the determined angle is used as a correction amount of the rotation angle, and the calculated angle between the two feature points is determined. The error in the distance between the camera and the recognition target is calculated from the distance between the two feature points determined by the distance and the position data, and the error is used as a correction value of the distance. Image recognition device as a feature.

(Supplementary Note 4) The image recognition means corrects the correction amount of the rotation angle or the correction value of the distance, and then determines the position of the characteristic point after the correction and the characteristic point determined by the position data. Supplementary notes 1 to 3 characterized in that the position is corrected.
The image recognition device according to any one of 1.

(Supplementary Note 5) Storage means for storing the template for the recognition object and the position data of the recognition object in the image of the camera, and pattern matching are performed in two different areas on one image, respectively. The position of the characteristic point is calculated in the region of, and the angle formed by the straight line passing through the two calculated characteristic points and the straight line passing through the two characteristic points determined by the position data is calculated, and the calculated angle is corrected by the rotation angle. A computer-readable recording medium having a program or a program for causing a computer to function as an image recognition means.

(Supplementary Note 6) Storage means for storing the template for the recognition object and the position data of the recognition object in the image of the camera, and pattern matching in two different areas on one image, The position of the feature point is calculated in each area, and from the distance between the two calculated feature points and the distance between the two feature points determined from the position data,
Calculating an error in the distance between the camera and the recognition object,
A computer-readable recording medium that stores a program for causing a computer to function as an image recognition unit that uses the error as a correction value for distance.

(Supplementary Note 7) Storage means for storing the template for the recognition object and the position data of the recognition object in the image of the camera, and pattern matching in two different areas on one image, The position of the characteristic point is calculated in each area, the angle formed by the straight line passing through the two calculated characteristic points and the straight line passing through the two characteristic points determined from the position data is calculated, and the calculated angle is corrected for the rotation angle. The amount of the error is calculated from the distance between the two characteristic points determined and the distance between the two characteristic points determined by the position data, and the error is calculated. A computer-readable recording medium having a program or a program for causing a computer to function as an image recognition unit that uses a distance correction value.

(Supplementary Note 8) After correcting the correction amount of the rotation angle or the correction value of the distance, the position of the characteristic point after the correction and the position of the characteristic point determined by the position data are corrected. A computer-readable recording medium recording the program for causing the computer according to any one of Supplementary notes 5 to 7 to function as an image recognition unit or the program according to any one of Supplementary notes 5 to 7.

[0095]

As described above, the present invention has the following effects.

(1): The image recognition means performs pattern matching in two different areas on one image, the position of the characteristic point is calculated in each area, and a straight line passing through the calculated two characteristic points is formed. Since the angle formed by the straight line passing through the two characteristic points determined by the position data is obtained and the obtained angle is used as the correction amount of the rotation angle, the correction amount of the rotation angle can be easily obtained from one image. .

(2): The image recognition means performs pattern matching in two different areas on one image, the position of the characteristic point is calculated in each area, and the distance between the two calculated characteristic points is calculated. To compare the distances between the two feature points determined from the position data, move the distance between the camera and the recognition object so that the compared distances are equal, and use the moved distance as a correction value of the distance. The correction value of the distance can be easily obtained from one image.

(3): The image recognition means performs pattern matching in two different areas on one image, the position of the characteristic point is determined in each area, and a straight line passing through the two determined characteristic points is obtained. An angle formed by a straight line passing through the two characteristic points, which is determined from the position data, is determined, the determined angle is used as a correction amount of the rotation angle, and the determined distance is determined from the distance between the two characteristic points and the position data. A distance between two characteristic points is compared, the distance between the camera and the recognition object is moved so that the compared distances are equal, and the moved distance is used as a correction value of the distance. Thus, the correction amount of the rotation angle and the correction value of the distance can be easily obtained.

(4): The image recognition means corrects the correction amount of the rotation angle or the correction value of the distance, and then corrects the position of the characteristic point after the correction and the position of the characteristic point determined by the position data. Since the correction is performed in the XY axis directions by eliminating the error of the rotation angle and the error of the distance in one image, the correction of the hand including the camera can be easily and accurately performed.

(5): Storage means for storing the template for the recognition object and the position data of the recognition object in the image of the camera, and pattern matching is performed in two different areas on one image, respectively. The position of the characteristic point is calculated in the region of, and the angle formed by the straight line passing through the two calculated characteristic points and the straight line passing through the two characteristic points determined by the position data is calculated, and the calculated angle is corrected by the rotation angle. Since the image recognition means is a program for causing a computer to function, by installing this program in the computer, an image recognition device that can easily obtain the correction amount of the rotation angle with one image is easy. Can be provided to.

[Brief description of drawings]

FIG. 1 is an explanatory diagram of an image recognition device of the present invention.

FIG. 2 is an explanatory diagram of a configuration of a substrate automatic check device according to an embodiment.

FIG. 3 is an explanatory diagram of a control block of the automatic board checking device according to the embodiment.

FIG. 4 is an explanatory diagram for inclining the probe pin according to the embodiment.

FIG. 5 is an explanatory diagram when one template is used in the embodiment.

FIG. 6 is an image recognition processing flow in the embodiment.

FIG. 7 is an explanatory diagram when two templates are used in the embodiment.

FIG. 8 is an image recognition processing flow in the embodiment.

FIG. 9 is an explanatory diagram of similarity addition according to the embodiment.

FIG. 10 is an explanatory diagram of correction of a rotation angle (θ3) in the embodiment.

FIG. 11 is an explanatory diagram of height correction according to the embodiment.

FIG. 12 is an explanatory diagram of XY correction and rotation angle correction in the embodiment.

FIG. 13 is an explanatory diagram of a conventional example.

[Explanation of symbols]

1a Object to be recognized 2 camera 4 probes 10 Control unit 11a drive means 13a Image recognition means 16a storage means

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁷ Identification code FI theme code (reference) G06T 7/60 G06T 7/60 150P // G01R 31/02 G01R 31/02 F term (reference) 2F065 AA01 AA06 AA32 BB05 CC25 FF04 JJ03 QQ38 UU05 2G014 AB59 AC09 AC12 5B057 AA03 DA07 DB02 DC02 DC08 DC33 5L096 CA02 DA02 FA66 FA67 FA69 JA03 JA09 JA18

Claims (5)

[Claims] 1. A camera for photographing an object to be recognized, a storage unit for storing a template for the object to be recognized and position data of the object to be recognized in an image of the camera, an image of the camera and the template. Image recognition means for performing pattern matching, wherein the image recognition means performs pattern matching in two different areas on one image, finds the position of the feature point in each area, and determines the two features An image recognition apparatus characterized in that an angle between a straight line passing through a point and a straight line passing through the two characteristic points determined by position data is obtained, and the obtained angle is used as a rotation angle correction amount. 2. A camera for photographing an object to be recognized, a storage unit for storing a template for the object to be recognized and position data of the object to be recognized in an image of the camera, and an image of the camera and the template. Image recognition means for performing pattern matching, wherein the image recognition means performs pattern matching in two different areas on one image, finds the position of the feature point in each area, and determines the two features An image characterized by calculating an error in the distance between the camera and the recognition object from the distance between the points and the distance between the two characteristic points determined by position data, and using the error as a correction value of the distance. Recognition device. 3. A camera for photographing an object to be recognized, a storage unit for storing a template for the object to be recognized and position data of the object to be recognized in an image of the camera, an image of the camera and the template. Image recognition means for performing pattern matching, wherein the image recognition means performs pattern matching in two different areas on one image, finds the position of the feature point in each area, and determines the two features An angle formed by a straight line passing through the point and a straight line passing through the two characteristic points determined by the position data is obtained, and the obtained angle is used as a correction amount of the rotation angle, and the distance between the calculated two characteristic points is calculated. From the distance between the two feature points determined from the position data,
Calculating an error in the distance between the camera and the recognition object,
An image recognition device characterized in that the error is used as a correction value of distance. 4. The image recognition means corrects the correction amount of the rotation angle or the correction value of the distance, and then determines the position of the feature point after the correction and the position of the feature point determined from the position data. The image recognition apparatus according to any one of claims 1 to 3, wherein correction is performed. 5. A storage unit for storing a template for a recognition target and position data of the recognition target in the image of the camera, and pattern matching in two different areas on one image, and each area. The position of the characteristic point is calculated with, and the angle formed by the straight line passing through the two calculated characteristic points and the straight line passing through the two characteristic points determined from the position data is calculated.
A program for causing a computer to function as an image recognition unit that uses the obtained angle as a correction amount of a rotation angle.