JP2613645B2 - Visual processing method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial applications]

The present invention relates to a visual processing method for recognizing an object having a complicated shape using a camera.

[Prior art]

Conventionally, when performing visual processing, circles have been used because of easy detection of feature amounts. In order to recognize a circle as a circle, we look at the character of the circle. The circularity is how close to a perfect circle is. Generally, the variation ratio of the outer circumference distance [determines the center of gravity of the circle, and determines the maximum value of the distance (radius) from the center of gravity to the outer circumference and the distance from the center of gravity to the outer circumference. The value obtained by dividing the difference between the minimum values of the distance (radius) by the average value of the distance (radius) from the center of gravity to the outer periphery] is closer to zero, it can be determined that the shape is closer to a perfect circle. Then, a method has been adopted in which the positions of the centers of gravity of the two circles are known, and a straight line connecting the centers of gravity is determined to determine the inclination of the target object.

[Problems to be solved by the invention]

Particularly, a target object such as a coarse material (casting) of an engine cylinder head having a complicated shape is also included in the target object.
Approximately three holes were imaged with a CCD camera, binarized, and recognized by extracting the roundness and color change. However, in a rough material, there are many burrs around a hole, and the burrs cause poor recognition of circularity. When the center (center of gravity) coordinate points of several holes are simultaneously searched, the center coordinate points are present at different angles obliquely to the optical axis of the CCD camera. Since the appearance of each hole differs due to the difference in the angle and the reflection angle of the proof light source also differs, it is often difficult to perform a binarization process in which the brightness differs around each hole. Also, it is very often difficult to perform a binarization process in which the color of the target object is slightly different for each part due to heat treatment or the like. Therefore, in the above case, a recognition (visual processing) error by the CCD camera occurs and visual processing cannot be performed. The present invention has been made in order to solve the above-described problems, and an object of the present invention is to move a camera with respect to a target object when visual recognition of the target object is poor in visual processing, and to set the visual angle thereof. It is to perform stable recognition of the target object by changing and recognizing the visual processing error factors.

[Means for Solving the Problems]

The configuration of the invention for solving the above-mentioned problem calculates a center of gravity of a reference shape provided on a target object from a binarized screen imaged by a camera, and moves the optical axis of the camera to the position of the center of gravity. In a binarized screen imaged by the camera at the position of the center of gravity, a measurement target shape having a predetermined feature amount is extracted from a region within a predetermined area range, and a region that is not recognized as a measurement target shape among the regions is extracted. In some cases,
The optical axis of the camera is moved to substantially the center between the position of the center of gravity and the unrecognized area, and the area on the binarized screen newly captured by the camera is used as a measurement target shape having the predetermined feature amount. Extract what is recognized,
The distance between the centers of gravity of the extracted measurement target shape and the reference shape is compared and determined using a preset value, and the inclination of the target object is calculated from the set of measurement target shapes specified in the comparison determination. It is characterized by the following.

[Action]

The target object is imaged by the camera, and the center of gravity of the reference shape provided on the target object is calculated from the binarized screen. Then, the optical axis of the camera is adjusted to the calculated center of gravity. Then, at this point, the positional relationship between the camera and the target object is clarified except for the angular deviation in the relative rotation direction. A measurement target shape having a predetermined feature amount is extracted from an area within a predetermined area range of the target object on the binarized screen of the camera at the position of the center of gravity of the reference shape. Of the regions that are not recognized as a measurement target shape, the camera moves the optical axis from the center of gravity of the reference shape to approximately the center of the region, and then performs a predetermined feature on a binarized screen in which the region is imaged again. An object which is recognized as a measurement target shape having an amount is extracted. As described above, by moving the optical axis of the camera closer to the unrecognizable area, the conditions in the visual processing of the camera are improved. That is, the angle between the optical axis of the camera and the region is reduced, and the uniformity of the illumination light source is achieved. As a result, the shapes to be measured can be extracted without omission, and further, the distance between the centers of gravity of the shapes to be measured and the reference shape is compared with a set value, and a set of shapes to be measured is recognized. Therefore,
The inclination of the target object is calculated by obtaining a straight line connecting the positions of the centers of gravity of the specified set of measurement target shapes, and at this point, the relative angular deviation between the camera and the target object in the rotational direction becomes clear.

【Example】

Hereinafter, a description will be given based on specific examples of the present invention. FIG. 1A and FIG. 1B are explanatory diagrams for explaining the visual processing method of the present embodiment. FIG. 1A shows a target object which is an engine cylinder head for visual processing.
11 shows a positional relationship between the CCD camera 12 and a total of 61440 pixels of 240 pixels in length and 256 pixels in width for measuring the position and inclination thereof. The engine cylinder head is made of a coarse material (cast), and generally has many burrs in the middle of the process and has color unevenness due to heat treatment or the like. FIG. 1 (b) shows a window 20 of the visual screen of the CCD camera 12, in which a reference hole 21 as a reference shape of the target object 11 and the optical axis of the CCD camera 12 are the center of gravity of the reference hole 21 of the target object 11. The position is shown. A group A composed of the reference hole 21 and the measurement target shape 22 of the target object 11 is shown on the left, and a B group composed of the reference hole 21 and the measurement target shape 23 of the target object 11 is shown on the right. . The measurement target shape 22 of the A group cannot be recognized at this camera position (broken line), and the measurement target shape 23 of the B group can be recognized at this camera position (solid line). The position where the optical axis of the CCD camera 12 is moved when the A group or the B group cannot be recognized is represented by A- or B-. Note that a control device (not shown) in the visual processing has an automatic filter function for improving recognition accuracy, a binarization level adjustment, a learning function associated with the level adjustment, and the like. Next, the visual processing method of the present embodiment will be described based on the flowcharts of FIGS. 2 (a) and 2 (b). FIG. 2A is a program for moving the optical axis of the CCD camera 12 to the reference hole 21. First, in step 100, the target object 11 is imaged in the window 20 of the visual screen of the CCD camera 12, and the binarized screen is captured. Next, the process proceeds to step 102, where a reference hole 21 corresponding to a preset value is extracted on the binarized screen of the captured target object 11. The reference hole 21 has a shape close to a perfect circle so that it can be extracted without fail even if the target object 11 is a coarse material (casting) such as an engine cylinder head, and is located substantially at the center of the target object 11. Holes have been selected. Next, the process proceeds to step 104, where the center of gravity of the reference hole 21 is calculated. Then, the process shifts to step 106 to shift the optical axis of the CCD camera 12 to the calculated position of the center of gravity, and ends the program. After the above-described program is executed, a program for calculating the inclination of the target object 11 shown in FIG. 2B is executed. In step 200, the window of the visual screen of the CCD camera 12
At 20, the target object 11 is imaged and its binarized screen is captured. Next, proceeding to step 202, of the area in the binarized screen of the target object 11 captured by the CCD camera 12 as the primary feature amount extraction, the area of the area of 12 pixels or less is deleted as noise, and is deleted. All the areas whose area is 13 pixels or more are extracted. Next, the process proceeds to step 204, where it is determined whether or not any of the extracted regions has an area exceeding 8000 pixels. That is, it is assumed that the area of the true region (measurement target shape) is preset to be 13 to 8000 pixels in terms of the number of pixels, and the area of the extracted region is also in the same pixel number range. Then, in step 204, if the area of the region extracted in step 202 overflows beyond 8000 pixels, it is determined as YES, and the process proceeds to step 206, where it is determined whether or not the number of measurements is less than three. You. If the number of times of measurement is less than three in step 206 and it is determined as YES, the process proceeds to step 208. Since the binarization level is not appropriate, the binarization level is changed, and the process returns to step 202 again. Step 2
When the number of times of measurement becomes three and the determination is NO in 06, it is determined that the measurement is impossible, and that the visual processing is NG is notified. If the area of the region extracted in step 204 does not exceed 8000 pixels, no overflow has occurred, so the determination is NO.
Then, the process proceeds to step 210, and as a secondary feature value extraction, for the region extracted by the temporary feature value extraction in step 202, the area of the region exists in a predetermined allowable area of the measurement target shape, A circumscribed rectangle composed of a main axis length and a sub axis length circumscribing the area of the target object 11 exists in an allowable range of a circumscribed rectangle composed of a main axis length and a sub axis length (vertical and horizontal lengths) of a predetermined measurement target shape The extracted area is extracted as a measurement target shape, and the position of the center of gravity of the extracted measurement target shape is calculated. Here, as described above, generally, if the area (measurement target shape) is a hole, it can be recognized by measuring the circularity. However, when the target object 11 is a coarse material (casting) and there are many burrs around the hole and it is far from a perfect circle, it is not possible to use the method of recognizing the circle to derive its center of gravity because of a large error. . Therefore, such a target object
In the case of 11, as shown in FIG. 1 (b), a circumscribed rectangle CR composed of a main axis length ML and a sub axis length SL circumscribing the region (measurement target shape) is determined in advance. A method of using the circumscribed rectangle CR based on the main axis length ML and the sub-axis length SL of the area (measurement target shape) to determine whether or not they can be recognized from upper and lower tolerances is effective because the error is reduced. Next, proceeding to step 212, it is determined whether or not there is an area that has been extracted at the time of step 204 and has been excluded in the secondary feature quantity extraction processing in step 210. If there is an excluded region, the determination in step 212 is Y
ES is reached, and the process proceeds to step 214 to move the optical axis of the CCD camera 12 to a substantially central position between the position of the center of gravity of the reference hole 21 and the position of the excluded region to be extracted. Next, proceeding to step 216, at the moved optical axis position of the CCD camera 12,
For the excluded area, the CCD camera 12 newly captures an image of the target object 11, captures the binarized screen, and performs the same processing as described in step 210. In other words, the extraction conditions of step 210 are applied again to exclude regions that do not meet the preset conditions, and further extract the conforming regions as measurement target shapes, and calculate their center of gravity positions. Next, proceeding to step 218, it is determined whether or not re-extraction has been performed with the optical axis of the CCD camera 12 moved for all the excluded regions. If there are a plurality of excluded regions in step 218, the determination is no, and the process proceeds to step 214, and steps up to step 218 are similarly repeated. In step 218, when the movement of the optical axis of the CCD camera 12 for re-extraction is completed for all the excluded regions, the determination is YES, and the process proceeds to step 220. Note that the determination is N when there is no area excluded in step 212 described above.
The process also proceeds to step 220 when the result becomes O. Then, in step 220, the number of extracted measurement target shapes is 2 to 20.
Is determined. Here, as a determination criterion in step 220, if it is sufficient that the number of extracted measurement target shapes is two or more, if an extremely large number of measurement target shapes are extracted, the subsequent processing time becomes a problem. If the number is set to about 2 to 6, the preset true measurement target shape may be excluded from the extracted measurement target shapes. Therefore, the upper limit of the number of extracted measurement target shapes is set to be too small. Is not good either. If the number of measurement target shapes extracted in step 220 is out of the setting range of 2 to 20, the result is NO, and step 222
Then, it is determined whether or not the time is over the preset time. If the time is not over in step 222, it is NO and the process proceeds to step 224,
Since the binarization level is not appropriate, the process proceeds to step 202 after the binarization level is changed, and the subsequent steps are the same as described above. Here, if the time is over in step 222 and YES is determined, it is determined that measurement is not possible and that visual processing is NG is notified. The above processing is performed, the number of measurement target shapes extracted in step 220 becomes 2 to 20, and if YES is determined, the process proceeds to step 226, and the measurement target shape used for determination in step 220, that is, From the plurality of extracted measurement target shapes, the first measurement target shape is taken out as the measurement target shape 1, and the distance between the centers of gravity between the measurement target shape 1 and the reference hole 21 is calculated. Next, the process proceeds to step 228, where it is determined whether the calculation of the distance between the centers of gravity has been completed for all of the plurality of extracted measurement target shapes. If the calculation has not been completed for all the measurement target shapes in step 228, the determination is NO.
Then, the process proceeds to step 230, in which the second measurement target shape is extracted from the plurality of extracted measurement target shapes, is set as the measurement target shape 1, and the process proceeds to step 226. The distance between the center of gravity and the reference hole 21 is calculated. In step 228, when the calculation of the distance between the centers of gravity is completed for all the extracted measurement target shapes, the determination is YES and the process proceeds to step 232. In step 232, it is determined whether or not one of the distances between the centers of gravity between the shape 1 to be measured and the reference hole 21 is within a preset distance tolerance.
In step 232, if there is no or more than two distances between the centers of gravity of the measurement target shape 1 and the reference hole 21 within the preset distance tolerance, the process proceeds to step 222, and thereafter, It is described in the same way. In step 232, it is determined that the distance between the centers of gravity of the measurement target shape 1 and the reference hole 21 is within a preset distance tolerance is one, and the determination is YES. Then, the measurement target shape used for the determination in step 220, that is, the first measurement target shape out of the plurality of extracted measurement target shapes is taken out as the measurement target shape 2, and the measurement target shape 2 The distance between the center of gravity and the center of gravity of the reference hole 21 is calculated. Next, the process proceeds to step 236, where it is determined whether the calculation of the distance between the centers of gravity has been completed for all of the plurality of extracted measurement target shapes. And
If the calculation of the distance between the centers of gravity is not completed for all the shapes to be measured in Step 236, the determination is NO, and Step 238 is performed.
Then, the second measurement target shape is taken out of the plurality of extracted measurement target shapes, is set as the measurement target shape 2, and the process proceeds to step 234. Similarly to the above, the measurement target shape 2 and the reference hole 21 are extracted. Is calculated. In step 236, when the calculation of the distance between the centers of gravity is completed for all the extracted shapes to be measured, the determination is YES and the process proceeds to step 240. In step 240, it is determined whether or not one of the distances between the centers of gravity between the measurement target shape 2 and the reference hole 21 is within a preset distance tolerance. Then, in step 240, the measurement target shape 2 and the reference hole 21
If there is no or more than two distances between the centers of gravity within the preset distance tolerance, the process proceeds to step 222, and the subsequent description will be made in the same manner as described above. If it is determined in step 240 that the distance between the centers of gravity between the measurement target shape 2 and the reference hole 21 is within a preset distance tolerance is one, the determination is YES, and after storing the measurement target shape 2, step 242 is performed. Then, the inclination of a straight line connecting the respective centers of gravity of the measurement target shape 1 stored in step 232 and the measurement target shape 2 stored in step 240 is calculated, so that the CCD camera and the target object This program in which the relative angular deviation in the rotational direction is determined is ended. As described above, first, the optical axis of the CCD camera 12 is
Is moved to the reference hole 21, and an area recognizable in the window 20 of the visual screen at that position is extracted as the shape to be measured. Then, if there is an unrecognizable area at that position, the optical axis of the CCD camera 12 is aligned with the reference hole 2 of the target object 11.
Move from 1 to the area, and extract the newly recognizable area as the measurement target shape. By performing such a visual processing method, the angle difference between the optical axis of the CCD camera 12 and the area (measurement target shape) of the target object 11 and the difference in the reflection state of the illumination light source are eliminated, and the target object 11 in the CCD camera 12 is eliminated. The extraction condition with the region (measurement target shape) is improved. Furthermore, the distance between the centers of gravity of the extracted measurement target shape and the reference hole is compared and determined by a preset value, and a specified set of measurement target shapes is recognized. In this way, the recognition (visual processing) error of the CCD camera 12 is absorbed, and the measurement target shape of the target object is stably recognized, so that the relative positional relationship between the CCD camera 12 and the target object 11 including the rotation direction is obtained. Is required.

【The invention's effect】

The present invention moves the optical axis of the camera to the position of the center of gravity calculated from the reference shape of the corresponding object imaged by the camera, and selects an area within a predetermined area range in the binarized screen imaged at that position. A measurement target shape having a predetermined characteristic amount is extracted, and the optical axis of the camera is moved to a position of the center of gravity and substantially the center of the region with respect to a region excluded from the measurement target shape, and the same extraction conditions are used. Since the shape to be measured is newly extracted in, the shape to be measured of the target object is extracted without omission, and further, the distance between the centers of gravity of the extracted shape to be measured and the reference shape is compared with a predetermined value and determined. As a result, a true set of measurement target shapes can be specified, so that the inclination of the target object can be reliably obtained.

[Brief description of the drawings]

FIG. 1A is a perspective view showing a positional relationship between a camera and a target object in a visual processing method according to a specific embodiment of the present invention. FIG. 1B is an explanatory diagram showing the appearance of the shape of the object to be measured on the visual screen of the camera in FIG. 1A. FIGS. 2 (a) and 2 (b) are flowcharts for achieving the visual processing method according to the embodiment. 11: target object, 12: CCD camera 20: window of visual screen, 21 …… Reference holes 22,23 …… Measurement target shape ML …… Spindle length, SL …… Counter shaft length, CR …… A circumscribed rectangle

Continued on the front page (72) Inventor Hisao Yamaguchi 1 Toyota Town, Toyota City, Aichi Prefecture Inside Toyota Motor Co., Ltd. (72) Inventor Yoshiji Yamamoto 1-1-1, Asahimachi, Kariya City, Aichi Prefecture Toyota Koki Co., Ltd. ( 56) References JP-A-61-213705 (JP, A)

Claims (1)

(57) [Claims] 1. A center of gravity of a reference shape provided on a target object is calculated from a binarized screen imaged by a camera, the optical axis of the camera is moved to the position of the center of gravity, and the camera is moved at the position of the center of gravity by the camera. In the imaged binarized screen, a measurement target shape having a predetermined feature amount is extracted from a region within a predetermined area range, and if there is any of the regions that is not confirmed as the measurement target shape, the camera Is moved to the approximate center of the position of the center of gravity and the unrecognized region, and the region on the binarized screen newly captured by the camera is recognized as the measurement target shape having the predetermined feature amount. The distance between the centers of gravity of the extracted measurement target shape and the reference shape is compared and determined by a preset value, and the object is determined from a set of measurement target shapes characterized by the comparison determination. The visual processing method characterized by calculating the slope of the body.