JPH08101913A - Method and device for indicating coordinate of circular inspection object area and inspection device using same device - Google Patents

Abstract

PURPOSE: To provide a coordinate indication method capable of performing coordinate access for performing a prescribed inspection by using an area (model picture) smaller than an inspection object area at a high speed. CONSTITUTION: For picture data obtained by picking up the image of an object K provided with a circular inspection object area, a window W in contact with the X axis of an orthogonal coordinate system is set, the circular inspection object area is parallelly moved and the center O is positioned at an original point (0, 0.) Also, it is rotated for a prescribed angle, an affine transformation parameter for which the picture data inside divided areas divided by each prescribed angle with the center of the circle as a reference are virtually present inside the window is used, affin transformation is performed to coordinate input in an access area and the picture data inside the corresponding prescribed divided area are accessed. Thus, for the instruction of coordinates, by specifying only the coordinate inside the window at all times and performing the affin transformation based on the entire angle θ, the entire inspection object area is made accessible.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a coordinate pointing method and apparatus for a circular inspection target area and an inspection apparatus using the apparatus.

[0002]

2. Description of the Related Art For example, in a visual inspection apparatus for inspecting a product for defects in appearance, an image of an object to be inspected is picked up by a camera, and the obtained image data is compared with a reference to be registered in advance. The image is compared (pattern recognition) to determine whether or not they match.

[0003]

By the way, the reference image registered as the comparison target has the same shape as the inspection target area, and the larger the inspection target area, the larger the reference image accordingly. As a result, a large memory capacity for storing the reference image is required.

[0004] Normally, the reference image to be registered captures a normal (non-defective) product (object), extracts an inspection target area in the captured image data, and registers it in a predetermined storage means. However, since such extraction specifies a rectangular region (n × m local region) including the inspection target region, for example, when the inspection target region is a ring, the central portion is for inspection. Although it becomes an unnecessary part as a model image, the memory is also used for that part, and the memory use efficiency is poor.

Furthermore, in the recognition process, it is only determined whether the object to be processed matches the reference image, and it is possible to determine which part in the inspection area does not match. There wasn't.

The present invention has been made in view of the above background, and an object of the present invention is to solve the above problems and perform a predetermined recognition by using a model image composed of an area smaller than an inspection target area. -Inspection processing can be performed, the memory efficiency of the reference image can be improved, and a position different from the reference image such as stains, chips, burrs on the object to be processed can be detected. And a method and apparatus for designating coordinates of a circular inspection target area, which has a simple control algorithm for reading data in image data captured for recognition / inspection processing and can be simplified in structure, An object is to provide an inspection device using the device.

[0007]

In order to achieve the above-mentioned object, in the coordinate pointing method for a circular inspection target area according to the present invention, image data obtained by imaging an object having a circular inspection target area. On the other hand, an access area that is parallel to any one of the orthogonal axes of the orthogonal coordinate system is set, and the circular inspection target area is at least rotationally moved to divide the circular inspection area at predetermined angles based on the center of the circle. Using the affine transformation parameter such that the image data in the divided area virtually exists in the access area, the affine transformation is performed on the coordinate input in the access area, and the corresponding image in the predetermined divided area I tried to access the data.

Preferably, one side of the access area is in contact with any one of the orthogonal axes, and the circular inspection area is translated so that the center of the circle is the origin of the orthogonal coordinate system. The access area is set to a predetermined position such that the divided area is virtually present in the access area when rotated at a predetermined angle while being positioned at
The affine transformation parameter is to determine that the image data in the divided area is virtually present in the access area by moving the circular inspection area in parallel and rotating.

In the pointing device according to the present invention suitable for carrying out the above method, an image memory for storing image data of an object having a circular inspection target area, and the circular inspection target area are stored. And an access means for accessing image data in a divided area obtained by dividing the circle by a predetermined angle with respect to the center of the circle, the access means using affine transformation, and the predetermined angle based on at least the center of the circle. And an affine transformation parameter for the access means so that the access area is parallel to one of the orthogonal axes of the orthogonal coordinate system.

Further, an image memory for storing image data of an object having a circular inspection target area, and a divided area obtained by dividing the circular inspection target area at a predetermined angle with the center of the circle as a reference. Access means for accessing the image data of, the access means uses affine transformation to translate the center of the circle in parallel to the origin position of the Cartesian coordinate system, and at the predetermined angle based on the center of the circle. It is more preferable to include a parameter setting unit for rotating the affine transformation parameter and setting the affine transformation parameter for the access unit so that the access region contacts one of the orthogonal axes of the orthogonal coordinate system.

In each of the above devices, the circular inspection target area may be an overlapping area of two circular windows having the same center, and a predetermined mask area may be added to the access area. You may

Further, in the inspection apparatus according to the present invention, any one of the above circular area coordinate pointing devices is provided, and a model memory having comparison reference data corresponding to the divided areas and predetermined coordinates in the access area are stored. The image data in the divided area in the image memory obtained by the access means based on the coordinate instructing means for instructing and the coordinates instructed by the coordinate instructing means is matched with the comparison reference data stored in the model memory. The recognition means is used.

[0013]

When the coordinates in the access area set parallel to one axis of the orthogonal coordinate system are sequentially designated, the affine transformation is performed based on the designated coordinates according to the predetermined affine transformation parameter. Then, the address of the predetermined divided area in the object stored in the image memory is output by the access means.

Therefore, when the image memory is accessed on the basis of the address, predetermined image data is read out. This is because the image of the object is rotated by a predetermined angle with the center of the circle as a reference. It is possible to virtually move the partial regions of the target object into the access region in order, and finally allow the entire inspection target region of the target object to virtually exist in the access region.

Further, since the access area is parallel to one of the coordinate systems, if all or predetermined coordinates in the access area are sequentially instructed, for example, in the raster direction, the access area will be Once the start point and its vertical and horizontal widths are known, the x and y coordinates can be easily incremented by one and repeated, and can be easily designated.

When the access area is moved so that the center of the circle is located at the origin of the Cartesian coordinate system, if the divided area is set to exist, the affine transformation parameters are simplified and one of them is set. Since the part becomes a constant, the coordinate designation can be performed faster.

Then, when the image data in the image memory is read according to the coordinate instruction and the matching is performed by the recognition means, it is determined whether or not the read divided area matches the comparison reference data. As a result, the inspection process accompanying the matching with the small model image can be performed on the entire inspection region. Then, since the read divided areas can be known from the rotation angle θ, the non-matching portion (defective portion) can be easily detected.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A preferred embodiment of a method and apparatus for pointing coordinates of a circular inspection target area according to the present invention and an inspection apparatus using the apparatus will be described in detail with reference to the accompanying drawings. The present invention relates to various devices and methods for performing recognition / inspection processing on an object having a circular (including a circle and a ring shape with a missing center) inspection object. In the present embodiment, a flat washer is used. Inspection device for an object having a ring-shaped area to be inspected, such as a peripheral portion such as the top surface of an object having a circular bottom surface or a top surface such as the whole or a bottle cap, a can for drinking water and canning, and mounting it The example applied to the coordinate pointing device of the circular area is shown.

The object will be described in more detail. For example, a flat washer or the bottom of a can of food, such as a flat washer or the like, has a uniform feature amount such as no pattern (the pattern is Even if it is not extracted as a feature amount), or even if the feature amount is not uniform over the entire circumference such as a gear, the same feature amount is repeated every fixed rotation angle. The object whose data (image data) appears is the object to be processed. If the inspection target area satisfies the above conditions, the shape of the target object does not necessarily have to be circular.

Next, the apparatus of the present embodiment will be described. As shown in FIG. 1, first, an object is imaged by the camera 1, and the obtained image (video signal) is A / D converter 2. After being converted into a digital signal, the image data is stored in the image memory 3, and the image data stored in the image memory 3 is D /
The A converter 4 converts the video signal again and outputs the video signal to the monitor 5. Then, the image is taken into the image memory 3 at a predetermined timing while the object is present in the image pickup area of the camera 1. As a result, for example, even if an object is randomly transported on the unloading side conveyor of the manufacturing line, by capturing image data at a certain moment, the monitor 5 displays a still image of the entire object. It will be output.

Here, in the present invention, based on a control signal from the CPU 6, the image data of a predetermined area stored in the image memory 3 is sent to the affine transformation section 8 as an access means, and a predetermined affine transformation is performed there. , The pixel data stored at a predetermined address is read, a predetermined feature amount is extracted by the matching processing unit 8, a predetermined recognition process such as comparison with a model image stored in the model memory 9 is performed, and The recognition result is output to the CPU 6.

Then, a ring-shaped object K is taken into the image memory 3 as shown in FIG. Further, a predetermined comparison reference value (numerical data such as average density and a model image M showing a part (for a predetermined angle) of the ring-shaped object K as shown in the figure) is stored in the model memory 9. ing. In the illustrated example, the model image M is enlarged for convenience.

Then, of the object K stored in the image memory 3, the image data existing in the window W, which is a rectangular access area of WX × WY from the coordinates (Xs, Ys) as a starting point, is read out and the matching processing section 8 is read. To obtain the density data of a predetermined pixel existing in the window, integrate the data, and then calculate the average density, and output the calculation result to the CPU 6. Also,
A predetermined pattern matching may be performed with the model image in addition to or instead of calculating the average density, and in such a case, the matching degree data is also output.
If the setting position of the window W is parallel to at least one of the X axis and the Y axis forming the Cartesian coordinate system, the installation position is not limited to the position at an angle of 0 degree as shown in the drawing Is.

Further, when obtaining the average density, a predetermined mask may be applied to the window W so that, for example, a portion corresponding to the region M1 other than the object in the model image is not read (not integrated). . The area M1 in which the mask is provided may be set to have substantially the same shape as the outer shape of the object when, for example, detecting stains or chips on the object. In the case of detection, a region that is slightly larger than the outer shape of the target object is set as the processing (integration) target region, and the change is appropriately performed based on the application of the image processing.

Further, by rotating the image of the object K by a predetermined angle with the center O of the circle as a reference, a partial area of the object K is virtually moved into the window W sequentially, and finally, The entire inspection area of the object K (the entire object in this example) is positioned within the window, and the model memory 9
Matches with the model image stored in. As a result, the inspection process accompanying the matching with the small model image can be performed on the entire inspection region.

The process of virtually moving each part of the object K sequentially into the window W is performed by coordinate conversion in accordance with an instruction of the address of the pixel to be read in the image memory from the affine conversion unit 7. I have to.

That is, assuming that the size of the image memory is, for example, the number of pixels in the X direction is Xsize and the number of pixels in the Y direction is Ysize as shown in FIG. 3, the upper left origin (0,0)
Letting VP be the start address in the image memory in which is registered, the address in the image memory (CPU memory space) where the coordinates (x, y) of an arbitrary point P in the image memory is stored is: address = VP + y × Xsize + x Has become.

Then, as shown in FIG. 4, a point P '(x', y ') on a certain circumference is defined by the center O (xc, yc) of the circle.
) Is rotated by an angle θ, and the coordinates of the point P ′ for making access at the coordinates (x, y) of the point P are:

[0029]

X ′ = xc + (x−xc) cos θ + (y−yc) sin θxc−xc cos θ−yc sin θ + xcos θ + ysin θ (1) y ′ = yc−xsin θ + xc sin θ + ycos θ−yc cos θ yc + xc sin θ−yc cos θ−xsin θ + y cos θ (2)

When RegA = xc-xc cos θ-yc sin θ RegB = yc + xc sin θ-yc cos θ RegC = cos θ RegS = sin θ, the above equations (1) and (2) are given by It can be replaced by the following formula.

[0031]

## EQU2 ## x '= RegA + x.RegC + y.RegS (3) y' = RegB-x.RegS + y.RegC (4), and the address in the CPU memory space where the pixel data of the point P is stored. ,

[0032]

## EQU3 ## Address = VP + y'Xsize + x '(5)

Then, the above RegA, RegB, Reg
C and RegS are the center coordinates (xc, yc in the image memory.
) Is fixed, so once the rotation angle θ is determined,
All become constants. Therefore, when the coordinates (x, y) of the point P in the window W are input to the equations (3) and (4), the coordinates (x ', y') of the corresponding point P'in the image memory can be obtained. Then, the coordinate (x ′, y ′) can be calculated by the equation (5).
When input to, the address in the image memory in which the grayscale data of the point P'is stored is obtained.

Then, in this example, as shown in FIG. 2, the calculation of RegA, RegB, RegC and RegS is performed by C.
This is performed by the PU 6, is set in the affine transformation unit 7 as an affine transformation parameter, and the coordinates (x, y) in the window W are specified by the data for determining the processing area from the CPU 6 (Xs which is the start point of the window, By supplying Ys and the window width (WX, WY) to the matching processing unit 8, the matching processing unit 8 sequentially supplies the coordinates (x, y) within the window to the affine transformation unit 7. The coordinate (x, y) instruction sets the window W in parallel with the X axis and the Y axis. Therefore, when scanning is performed in the raster direction, x, y does not need special coordinate conversion. Can be easily instructed by appropriately incrementing and resetting.

Then, in the affine transformation unit 7, the given parameters RegA, RegB, RegC, RegS are given.
Then, based on the coordinates (x, y), the equations (3) to (5) are sequentially executed to calculate the address of the corresponding pixel in the image memory. Then, data is read from the image memory 3 based on the calculated address,
The data is sent to the matching processing unit 8. That is,
The affine transformation unit 7 is configured by appropriately combining an adder and a multiplier that execute the above equations.

Then, when performing the matching process on the object K stored in the image memory shown in FIG. 2, the predetermined angle θ is incremented from θs to θe at predetermined intervals dθ, and the window W is updated each time. By sequentially designating the coordinates (x, y) in the raster direction from the upper left start coordinate of the window, the image data of a predetermined area (shown by hatching in the figure) stored in the corresponding image memory 3 is displayed. It will read out and compare it with the model image.

In this example, since the average density of the predetermined area in the image memory is obtained, the model memory stores the widths WX and WY of the windows and the average density as a reference. , Window width W
After the CPU 6 once receives X and WY, they are given to the matching processing unit 8, and the reference average density is the CPU 6.
And the matching processing unit 8 obtains only the average density in the image memory and gives the calculation result to the CPU 6,
At U6, the calculation result is compared with the reference average density obtained from the model memory, and it is determined whether or not the difference is within the allowable range.

As the processing in the matching processing unit, the average density is obtained in this example, but various kinds of concrete matching (recognition processing) can be used. Image data (model image) to be stored in advance, and the matching processing unit 8
When the matching degree (fitness / similarity) of the image data in the image memory with respect to the model image is obtained, the model image is sent to the matching processing unit 8.
In the matching processing unit 8, the obtained matching degree and the like are calculated by the CPU.
6, the CPU 6 determines whether or not the degree of coincidence is greater than or equal to a predetermined threshold value, and the comparison criteria stored in the model memory 7 or the data transmission / reception path depending on the algorithm of the matching process used. The data also changes accordingly.

Then, in the concrete processing of the CPU 6 at that time, that is, one embodiment of the instruction method according to the present invention, as shown in the flowchart of FIG. Is specified, the center coordinates (xc, yc) thereof are calculated, and the start point (Xs, Ys) of the window is calculated based on the calculated center coordinates (xc, yc). That is, the center coordinates are
For example, it can be determined by detecting the inner circumference or the outer circumference of the object K based on an edge extraction algorithm and determining the midpoint of the diameter. For the diameter, for example, the interval between the pair of edges (horizontal both ends of the circle) detected when scanning in the X-axis direction is obtained, and when the value is the maximum, the diameter is the diameter of the circle. You can judge. In addition, although the detection of the center is operated by the software by the CPU in this example, it may be configured by hardware, and the algorithm for the detection is not limited to the above-described one, and various measures are taken. be able to.

Since the size and shape of the object (inspection area) is known, the coordinates (Xs, Ys) can be easily measured from the center.
) Can be calculated, and the window widths WX and WY are set in the matching processing section 8 together with the obtained (Xs, Ys) (ST1).

Next, n is reset, the rotation angle θ is determined (ST2 to ST3), four affine transformation parameters are obtained based on the obtained θ, and these are set in the affine transformation unit 7 (ST4). .

Then, based on each value obtained as described above and the coordinates (x, y) of each pixel in the window sent from the matching processing unit 8 to the affine transformation unit 7, the affine transformation unit 7 obtains the value. According to the corresponding address in the corresponding image memory, the data stored at that address is read and given to the matching processing unit 8. Then, the matching processing unit 8 integrates the data of the given pixels and obtains an average density by dividing the data by the number of times of integration, and all the target pixels in the window (entire window or a predetermined mask is applied). If the average density calculation is completed for the pixels existing in the area excluding the portion), the CPU 6 receives the calculation result (ST5). Each coordinate (x, y) in the window
Of course, the designation may be given from the CPU 6.

Then, the CPU 6 determines whether or not the given average density is within a certain range with respect to a predetermined reference value (good or bad). When the matching process for one area is completed, it is determined whether the currently processed rotation angle θ is the end angle, and if there is an unprocessed area (less than the end angle), θ is set. After incrementing by the angular interval dθ (ST6, 7, 3), the processing from step 4 onward is executed. Then, if the processing is performed for all the angles, the processing ends.

In this way, using a small model image, an inspection process can be performed on an object having a large circular inspection area. Moreover, since the degree of coincidence is partially inspected, if a region (defective portion) different from the model image is detected, it is possible to easily extract which portion of the object has a defect from the angle θ at that time. You can In the present embodiment, the CPU 6 constitutes a parameter setting means, and the matching processing section 8 constitutes a recognition means and a coordinate designating means.

On the other hand, as shown in FIG. 6, the point P '(x', y ') on a certain circumference is the center O (xc, yc) of the circle.
Is moved by an angle θ around the center, and then the coordinates (x,
The coordinates of the point P'to make it accessible in y) are

[0046]

(4) x '= xc + (x-xc-dx) cos θ + (y-yc-dy) sin θ = xc- (xc + dx) cos θ- (yc + dy) sin θ + xcos θ + ysin θ (6)

[0047]

Y '= yc- (x-xc-dx) sin θ + (y-yc-dy) cos θ = yc- (xc + dx) cos θ- (yc + dy) sin θ-xsin θ + ycos θ (7) ), And the address in the CPU memory space where the data of the pixel at the point P is stored is

[0048]

## EQU6 ## Address = VP + y'Xsize + x '(8) In the above relational expression, first, the center O (xc, yc) of the circle is moved in parallel by dx, dy by the parallel movement, and then the center of the moved circle is set as the rotation center, and the predetermined angle θ is obtained.
The same is true when only rotating.

Then, as shown in FIG. 7, the image memory 3
Matching is performed by moving the center O (xc, yc) of the object K stored in the virtual image to the origin (0,0) position at the upper left of the image memory and rotating it by a predetermined angle θ. The image data virtually moved into the processing area (window W) is compared with the model image.

Then, the absolute value of the moving distance dx in the X-axis direction is xc which is the X coordinate of the center O, and the absolute value of the moving distance dy in the Y-axis direction is yc which is the Y coordinate of the center O. Since the moving directions are both negative, therefore, dx and dy in the above equations (6) and (7) are respectively −
Substituting xc and -yc,

[0051]

X ′ = xc− (xc + (− xc) cos θ− (yc + (− yc) sin θ + xcos θ + ysin θ = xc + xcos θ + ysin θ ... (9)

[0052]

Y ′ = yc− (xc + (− xc)) cos θ− (yc + (− yc)) sin θ−xsin θ + ycos θ = yc−xsin θ + ycos θ ... (10) An arithmetic expression is obtained. Then, as in the case of only rotating the above equations (9) and (10), each parameter when replaced by the affine transformation parameter Reg is: RegA = xc RegB = yc RegC = cos θ RegS = sin θ.

Then, similarly to the above, the predetermined angle θ is set to θ
From s to θe, increment by a predetermined interval dθ,
Each time, the coordinate (x, y) in the window W is sequentially designated in the raster direction from the start coordinate at the upper left of the window, for example, so that a predetermined area stored in the corresponding image memory 3 (shown by hatching in the figure). Considering the processing of reading the image data of 1 and comparing the model images, the processing function of the CPU 6 in that case, that is, another embodiment of the instruction method according to the present invention, is as shown in the flowchart of FIG.

That is, first, in this example, the position of the window W is always constant regardless of the position of the object K in the image memory 3, so that the starting point (X ') of the window is set.
s, Y's) and window widths WX and WY are set in the matching processing unit 8 (ST11).

Next, the existence position of the object in the imaged image memory is specified, its center coordinates (xc, yc) are determined, and RegA, R which is one of the affine transformation parameters.
EgB is obtained and set in the affine transformation unit 8 (S
T12). Then, n is reset, the rotation angle θ is determined (ST13 to ST14), and the remaining two affine transformation parameters (Reg
C, RegS) is obtained and set in the affine transformation unit 7 (ST15).

Then, based on each value obtained as described above and the coordinates (x, y) of each pixel in the window sent from the matching processing unit 8 to the affine transformation unit 7, the affine transformation unit 7 obtains the value. According to the corresponding address in the corresponding image memory, the data stored at that address is read and given to the matching processing unit 8. Then, the matching processing unit 8 integrates the data of the given pixels and obtains an average density by dividing the data by the number of times of integration, and all the target pixels in the window (entire window or a predetermined mask is applied). If the average density calculation is completed for the pixels existing in the area excluding the portion), the CPU 6 receives the calculation result (ST16).

Then, the CPU 6 determines whether or not the average density is within a certain range. When the matching process for one area is completed, it is determined whether or not the currently processed rotation angle θ is the end angle. If there is an unprocessed area (less than the end angle), θ is incremented by the angular interval dθ (ST17, 1
8 and 14), the processes after step 15 are executed. Then, if the processing is performed for all the angles, the processing ends.

In the present embodiment, A and B of the affine transformation parameters are constants that do not depend on the rotation angle θ, so that as is clear from FIG.
Once gB is obtained, the subsequent calculation in CPU 6 is Re
Since only gC and RegS are used, the number of steps for performing arithmetic processing can be reduced, and since RegA and RegB need only be substituted with xc and yc, the arithmetic processing of the CPU 6 in the case of rotating on the spot described above. Compared with (Fig. 5), it can be processed at extremely high speed. as a result,
Inspection processing can also be performed at high speed. Note that the other configurations and effects are the same as those of the above-described embodiment, so detailed description thereof will be omitted.

In each of the above-described embodiments, an example in which the inspection target area has a ring shape has been described.
The present invention is not limited to this, and may be a circle. Then, in the case of a circle, for example, a window W, which is a processing area, may be set for the object K as shown in FIG. 9 and the matching processing may be performed for all of the windows W, or By masking the predetermined area M1 so as not to be a matching processing target, the matching processing may be performed based on a fan-shaped processing area shown by hatching in the drawing, and various modifications can be implemented. Is. Then, in the example shown in FIG. 9, the entire surface of the object K is the inspection target area.

[0060]

As described above, in the coordinate pointing method and apparatus for a circular inspection target area according to the present invention and the inspection apparatus using the apparatus, in the access area set parallel to one axis of the orthogonal coordinate system. When the coordinates are sequentially designated, the affine transformation is performed based on the designated affine transformation parameters, and the address of the predetermined divided area in the object stored in the image memory can be output by the access means. That is, instead of actually rotating the image data and the model, the image is read in the same posture as the image data and the like virtually stored in the model memory by appropriately converting the image when reading the image. You can Therefore, high-speed processing becomes possible.

Moreover, when the image memory is accessed based on the address, predetermined image data is read out,
This is because the image of the object is rotated by a predetermined angle with respect to the center of the circle to virtually move the partial areas of the object into the access area in order, and finally the object. Since the entire inspection target area can virtually exist in the access area, a predetermined recognition / inspection process can be performed using a model image including an area smaller than the inspection target area, Improve memory usage efficiency.

Further, since the access area is parallel to one of the coordinate systems, the instruction of all or predetermined coordinates in the access area by the coordinate instruction means can be obtained by a simple calculation process, which is simple. And can be instructed quickly.

Further, when the access area is moved so that the center of the circle is located at the origin of the Cartesian coordinate system, if the divided area is set to exist, the affine transformation parameters are uni-ordered and a part of them is converted. Since it becomes a constant, the number of steps of the arithmetic processing when changing the rotation angle is small,
In addition, since the calculation is simple, the coordinates can be designated at a higher speed.

Then, when the image data in the image memory is read according to the coordinate instruction and matching is performed by the recognition means, it is judged whether or not the read divided area matches the comparison reference data. Is a unit of divided area, and therefore it is possible to easily detect that the divided area is defective when it is determined that they do not match. That is, it is possible to detect a position different from the reference image such as dirt, chipping, or burr on the object to be processed.

[Brief description of drawings]

FIG. 1 is a diagram showing a preferred embodiment of the present invention.

FIG. 2 is a block configuration diagram showing a main part thereof.

FIG. 3 is a diagram illustrating an operation principle.

FIG. 4 is a diagram illustrating an operation principle.

FIG. 5 is a flowchart illustrating the functions of the CPU in this embodiment.

FIG. 6 is a diagram illustrating the operating principle of another embodiment.

FIG. 7 is a diagram illustrating the operating principle of another embodiment.

FIG. 8 is a flowchart illustrating the functions of the CPU in this embodiment.

FIG. 9 is a diagram illustrating another inspection target area.

[Explanation of symbols]

 1 camera 3 image memory 6 CPU (parameter setting means) 7 affine transformation section (access means) 8 matching processing section (coordinate pointing means, recognition means) 9 model memory W window (access area)

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification number Internal reference number FI Technical display location G06T 3/60 G06F 15/64 340 B 15/66 350 A

Claims (7)

[Claims] 1. An image memory for storing image data in which an object having a circular inspection target area is imaged, and a divided area obtained by dividing the circular inspection target area at a predetermined angle based on the center of the circle. Access means for accessing the image data of, the access means, using the affine transformation, to rotate at every predetermined angle with respect to at least the center of the circle, the access region is one of the orthogonal axes of the orthogonal coordinate system A coordinate pointing device for a circular area, comprising: parameter setting means for setting an affine transformation parameter for the access means so as to be parallel to the one side. 2. An image memory for storing image data in which an object having a circular inspection target area is imaged, and a divided area obtained by dividing the circular inspection target area at a predetermined angle based on the center of the circle. And an access unit for accessing the image data, wherein the access unit uses an affine transformation to move the center of the circle in parallel to the origin position of the orthogonal coordinate system,
And a parameter setting means for rotating the circular center with respect to each of the predetermined angles and setting an affine transformation parameter for the access means so that the access region contacts one of the orthogonal axes of the orthogonal coordinate system. Coordinate pointing device for circular area. 3. The coordinate pointing device for a circular area according to claim 1, wherein the circular inspection target area is an overlapping area of two circular windows having the same center. 4. The circular area coordinate pointing device according to claim 1, wherein a predetermined mask area is added to the access area. 5. The circular area coordinate pointing device according to claim 1, further comprising a model memory having comparison reference data corresponding to the divided area, and pointing a predetermined coordinate in the access area. And the image data in the divided area in the image memory obtained by the access means based on the coordinates instructed by the coordinate instructing means and the comparison reference data stored in the model memory are matched. An inspection device for a circular area, comprising a recognition means. 6. An access area parallel to any one of orthogonal axes of an orthogonal coordinate system is set for image data obtained by imaging an object having a circular inspection area, The access is performed by using an affine transformation parameter such that the image data in the divided area obtained by dividing the inspection area at least by rotating the inspection area at a predetermined angle based on the center of the circle is virtually present in the access area. A coordinate designating method for a circular area, wherein affine transformation is performed on a coordinate input in the area to access image data in the corresponding predetermined divided area. 7. One side of the access area is in contact with any one of the orthogonal axes, and the circular inspection target area is translated so that the center of the circle is located at the origin of the orthogonal coordinate system. And when rotated by a predetermined angle, the access area is set to a predetermined position such that the divided area virtually exists in the access area, and the affine transformation parameter moves the circular inspection target area in parallel and 7. The method for instructing coordinates of a circular area according to claim 6, wherein the image data in the divided area is determined to virtually exist in the access area by rotating and moving.