JP7259308B2 - Part shape inspection method - Google Patents

Description

The present invention relates to a shape inspection method for parts.

Conventionally, various component shape inspection methods have been used. For example, Patent Literature 1 discloses a gear shape inspection method in which image data of the entire gear to be inspected is compared with CAD data of the entire gear. Further, Patent Document 2 discloses a method for calculating the number of teeth of a gear using a pattern image. Further, Patent Document 3 discloses a method for extracting edges of pattern elements on a substrate.

JP 2003-166814 A JP-A-2001-99626 Japanese Unexamined Patent Application Publication No. 2002-116007

However, in the gear shape inspection method described in Patent Document 1, even if the gear meets the standard as a whole from a local point of view such as the shape and pitch of the teeth of the gear and is originally usable, the gear Comparing the gears as a whole causes the distance between both ends of the gear to increase, and judging the other end according to the one end may deviate from the standard and may be judged to be unusable. In addition, since the data of the entire gear is required, the amount of data increases, and the inspection load is heavy. Further, the method for calculating the number of teeth described in Patent Literature 2 uses a pattern image, so the amount of data is large, and the inspection load is heavy. The edge extraction method described in Patent Document 3 also increases the amount of data and the load of inspection. As described above, it is difficult for the conventional component shape inspection method to easily inspect the component shape with high accuracy.

A method for inspecting the shape of a component according to the present invention for solving the above-mentioned problems includes a comparison point cloud data reading step of reading point cloud data, which is comparison data corresponding to a part of the contour of the component, as comparison point cloud data. an image data acquisition step of acquiring image data of the component; a corresponding data acquisition step of acquiring corresponding data corresponding to the comparison data from the image data of the component; and image points that are point cloud data from the corresponding data. An image point cloud data extraction step of extracting group data, and a determination step of comparing the image point cloud data with the comparison point cloud data to determine the quality of the shape of the part.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a schematic diagram showing an example of a system capable of executing the component shape inspection method of the present invention; FIG. 4 is an image displayed on a monitor, which is a schematic diagram showing an example of a part whose shape is inspected by the method for inspecting the shape of a part according to an embodiment of the present invention; 4 is a flow chart of a component shape inspection method according to an embodiment of the present invention. FIG. 5 is an image displayed on a monitor, which is a schematic diagram showing an example of comparison data of parts according to one embodiment of the present invention. FIG. 4 is an image displayed on a monitor, which is a schematic diagram showing an example of comparison point cloud data used in the component shape inspection method according to one embodiment of the present invention. FIG. 4 is an image displayed on a monitor, which is a schematic diagram showing a circular image used in the component shape inspection method according to one embodiment of the present invention. 4 is an image displayed on a monitor, and is a graph showing a circular image as a strip image used in the component shape inspection method according to the embodiment of the present invention. An image displayed on a monitor, which is a band-shaped image used in a component shape inspection method according to an embodiment of the present invention, and which is a band-shaped image in the case where the center position is shifted and after the center position is adjusted. A graph representing a strip image of .

First, the present invention will be generally described.
According to a first aspect of the present invention, there is provided a component shape inspection method for solving the above-described problems. a group data reading step; an image data obtaining step of obtaining image data of the component; a corresponding data obtaining step of obtaining corresponding data corresponding to the comparison data from the image data of the component; An image point cloud data extraction step of extracting image point cloud data, which is data, and a judgment step of comparing the image point cloud data with the comparison point cloud data to judge the quality of the shape of the part. Characterized by

According to this aspect, it is possible to inspect the shape of the part with high accuracy by judging whether the shape of the part is good or bad based on the range corresponding to a part of the outline of the part. This is because the outline of the part is a part that greatly affects the performance of the part and has a clear shape. In addition, instead of comparing the images of the entire part, the images of a part of the outline of the part are compared, and the data of the point groups are compared with each other. Therefore, it is possible to prevent the amount of data from becoming too large and reduce the inspection load. That is, according to this aspect, the shape of the component can be easily inspected with high accuracy.

A second aspect of the present invention is a part shape inspection method according to the first aspect, wherein the part is a gear, and the range corresponding to a part of the contour of the part includes at least one tooth of the gear. characterized by comprising

Gears as parts are often used in large quantities in various devices, but according to this aspect, the shape of gears can be easily inspected with high accuracy.

A third aspect of the present invention is a component shape inspection method according to the first or second aspect, wherein the number of comparison point cloud data is ten times or more that of the image point cloud data. and

According to this aspect, the number of comparison point cloud data is 10 times or more that of the image point cloud data. The shape of the part can be inspected.

A fourth aspect of the present invention is a component shape inspection method according to any one of the first to third aspects, wherein the comparison point cloud data has normal line information, and the comparison point cloud data reading step is characterized by reading the position of the outline of the part based on the normal line information.

Judging the data along the normal increases the accuracy of data judgment. According to this aspect, the comparison point cloud data has normal information, and the position of the outline of the part is determined based on the normal information. , the data can be determined along the normal, and the shape of the part can be inspected with particularly high accuracy.

A fifth aspect of the present invention is a component shape inspection method according to any one of the first to fourth aspects, wherein the image point cloud data extraction step extracts the image point cloud data in units of pixels. A first image point cloud data extraction step and a second image point cloud data extraction step of extracting the image point cloud data in units of sub-pixels.

According to this aspect, it has a first image point cloud data extraction step of extracting image point cloud data in units of pixels, and a second image point cloud data extraction step of extracting image point cloud data in units of subpixels. That is, by executing rough inspection and detailed inspection, the shape of the part can be inspected quickly and with high accuracy.

A sixth aspect of the present invention is a component shape inspection method according to any one of the first to fifth aspects, wherein the image point cloud data extracted in the image point cloud data extraction step is used in the determination step. It is characterized by having an unused data extracting step of extracting unused data.

According to this aspect, it is possible to extract the noise data that lowers the accuracy of determining the quality of the shape of the component as unused data that is not used in the determination process, so that the shape of the component can be inspected with particularly high accuracy.

A component shape inspection method according to a seventh aspect of the present invention is, in any one of the first to sixth aspects, a data conversion step of converting the image point cloud data extracted in the image point cloud data extraction step. wherein the determining step uses the data converted in the data converting step to determine whether the shape of the component is good or bad.

According to this aspect, since the data conversion step of converting the image point cloud data is provided, for example, by converting a circular image into a band-shaped image, a plurality of similar shaped portions can be compared quickly and with high accuracy, and the center It is possible to correct the center position of the image data acquired with the position shifted, and it is possible to inspect the shape of the part with particularly high accuracy.

A method for inspecting the shape of a part according to an embodiment of the present invention will be described in detail below with reference to the accompanying drawings.

<Part shape inspection system>
First, a component shape inspection system 1 as an embodiment capable of executing the component shape inspection method of the present invention will be described with reference to FIGS. 1 and 2. FIG.

As shown in FIG. 1, the shape inspection system 1 of this embodiment includes a PC 2 having a monitor 2a and a storage unit 2b, a camera 3 having a telecentric lens 4, and parts such as a gear 5 to be measured. and a lighting device 7 for irradiating the stage 6 with light. Here, the camera 3, the stage 6 and the lighting device 7 are connected to the PC2, and the operator can operate the camera 3, the stage 6 and the lighting device 7 using software installed on the PC2.

The light emitted from the illumination device 7 can pass through the stage 6, and the camera 3 can capture an image as shown in FIG. 2, for example. The PC 2 is configured to be able to acquire the image captured by the camera 3 as image data. In addition, since the PC 2 has a storage unit 2b such as a hard disk or RAM, it is configured so that the operator can store comparison data in the storage unit 2b for comparing the quality of the shape of the part. there is CAD data or the like can be used as the comparison data.

Here, although the stage 6 of this embodiment is made of transparent glass, it does not have to be made of glass as long as it is configured to transmit the light emitted from the illumination device 7, and it may be made of resin. It is possible. In addition, the camera 3 of the present embodiment is a camera capable of photographing a monochrome image of 12 million pixels, and is capable of photographing an image with little distortion by being provided with the telecentric lens 4. There are no restrictions. In addition, although the illumination device 7 of this embodiment is configured to irradiate light from the side facing the camera 3, it is also possible to use a configuration that irradiates the stage 6 with light from the camera 3 side. It is possible.

<Part shape inspection method>
Next, a method for inspecting the shape of a part according to the present embodiment will be described with reference to FIG. 2 and FIGS. 4 to 8 using the flow chart of FIG. In the method for inspecting the shape of a component according to the present embodiment, the gear 5 for a wrist watch is used as the component. parts can be inspected.

When the component shape inspection method of the present embodiment is started, first, in step S110, from the CAD data, which is comparison data as shown in FIG. Load data. 4 shows CAD data of the gear 5 displayed on the monitor 2a. The CAD data of this embodiment is data in DXF format, but the data format is not particularly limited. In the CAD data of the present embodiment, line portions such as the contour of the tooth 5a of the gear 5 are vector data, and the vector data is converted into point cloud data to obtain comparison point cloud data. In this embodiment, an error on the order of several μm is measured, and in order to measure the error on this order, the comparison point cloud data is converted so as to have an accuracy of 0.1 μm.

The comparison point cloud data read in step S110 is represented by three types of models displayed on the monitor 2a shown in FIG. , point cloud data B for two adjacent teeth for pitch determination of the adjacent tooth 5a, or point cloud data C for one tooth and half teeth on both sides adjacent to the tooth for one tooth. .

Each piece of data forming the comparison point cloud data has normal line information. That is, it has information on the direction orthogonal to the vector direction of the vector data before conversion. By having normal line information and determining the position of the edge of the tooth 5a, which is the contour of the gear 5, in the direction along the normal line, the position of the edge of the tooth 5a can be determined with high accuracy.

Next, in step S120, the camera 3 captures an image of the gear 5 placed on the stage 6 to obtain image data. The image data corresponds to FIG. In this embodiment, image data of three gears 5 are acquired in this step.

Next, in step S130, pattern search is performed on the image data obtained in step S120. The pattern search is performed for each gear 5. FIG. Specifically, the pattern for pattern search is a grayscale image corresponding to one gear stored in advance in the storage unit 2b, and the pattern search is performed by searching for image data that matches the pattern. In this step, since the pattern or the image data can be rotated and searched, the search can be performed with high accuracy no matter what angle the gear 5 is placed on the stage 6 in the direction of rotation. can. In this embodiment, three gears 5 represented in FIG. 2 are searched in this step.

Next, in step S140, in the image data of the first gear 5, corresponding data corresponding to the comparison point cloud data is obtained. In this step, the data for the entire circumference of the gear 5 can be used as the corresponding data, but in this embodiment, only the arc area from the start point P1 to the end point P2 in FIG. 6 is designated as the target. are doing. In this embodiment, the start point P1 and the end point P2 are set to be the most recessed portion of the root portion of the tooth 5a so as to facilitate subsequent processing, but the positions of the start point P1 and the end point P2 are not particularly limited.

Next, in step S150, image point cloud data on the contour of the gear 5 is extracted from the corresponding data acquired in step S140. In this embodiment, the extraction of the image point cloud data is performed by extracting the image point cloud data at high speed in units of pixels using an algorithm based on the Canny method. However, there is no particular limitation on the method of extracting the image point cloud data.

Next, in step S160, in order to remove noise as unused data, the image point cloud data is labeled and grouped, and the image point cloud data of the group with the largest area among the grouped groups is Image point cloud data representing the contour of the gear 5 is used. That is, data deviating from the image point cloud data is removed as noise and becomes unused data.

Next, in step S170, the image point cloud data representing the contour of the gear 5 obtained in step S160 is converted into data of the distance seen from the center 5c of the gear 5 shown in FIG. The angle information of the group data is sorted counterclockwise and converted into a peak-and-valley profile as shown in FIG. That is, the circular image is converted into a band-shaped image by converting the image point cloud data of the gear 5 positioned substantially on the circumference from the polar coordinates to the orthogonal coordinates. By the data conversion in this step, in addition to extracting image point cloud data for one tooth, image point cloud data for two adjacent teeth are extracted, and half teeth on both sides adjacent to the tooth for one tooth are extracted. Image point cloud data can be extracted minute by minute.

Next, in step S180, in the image data of the gear 5, it is determined whether or not the position of the center 5c of the gear 5 is aligned. If it is determined that the position of 5c is not correct, the process proceeds to step S190. When the position of the center 5c is matched, the peak-and-valley profile as represented by the solid lines in FIGS. 7 and 8 is obtained. On the other hand, when the position of the center 5c is not aligned, the peak-and-valley profile as indicated by the dashed line in FIG. 8 is obtained.

In step S190, for example, the maximum peak point group is extracted from the image point cloud data representing the contour of the gear 5, circle approximation is performed using the maximum peak point group, and the center is set as the new center. provides center alignment. After execution of step S190, the process returns to step S140. After execution of step S190, in the data conversion of step S170, the peak-and-valley profile represented by the dashed line in FIG. 8 changes to the peak-and-valley profile represented by the solid line in FIG.

In step S200, in order to remove noise, low-pass processing is performed for filtering according to whether the pattern in the image is coarse or fine, and peak points are extracted from the peak-and-valley profile as shown in FIG. Extraction of peak points is performed, for example, by extracting points exceeding and falling below the average value of the image point cloud data, and extracting the maximum point between the points exceeding and falling below the average value. is extracted as the maximum peak point, and the point of the minimum value between the point exceeding the average value and the point falling below the average value is extracted as the minimum peak point.

Next, in step S210, the image point cloud data of adjacent minimum peak point-maximum peak point-minimum peak point is regarded as image point cloud data corresponding to one tooth 5a, and image point cloud data for each tooth 5a is generated. Extract. The image point cloud data extracted in this step corresponds to image point cloud data for one tooth for shape determination of each tooth 5a.

Next, in step S220, by extracting image point cloud data for two teeth from the adjacent minimum peak point-maximum peak point-minimum peak point-maximum peak point-minimum peak point, the adjacent peak point for pitch determination is extracted. Image point cloud data for two teeth are extracted. Alternatively, the image point cloud data for one tooth from the adjacent maximum peak point-minimum peak point-maximum peak point-minimum peak point-maximum peak point and half teeth on both sides adjacent to the one tooth. By extracting, image point cloud data for one tooth for pitch determination and half teeth on both sides adjacent to the one tooth are extracted.

Next, in step S230, the comparison point cloud data in step S110 and the image point cloud data for shape determination of each tooth 5a in step S210 are subjected to primary matching for each tooth 5a. As a matching method, for example, a matching method using ICP (Iterative Closest Point) can be executed. When ICP is used, if the initial positions of the comparison point cloud data and the image point cloud data are largely misaligned, the results do not converge. Matching is performed 12 times while shifting by 30 degrees, and the result with the best evaluation score can be selected. Then, image point cloud data corresponding to the comparison point cloud data is extracted, and normal line information of the corresponding comparison point cloud data is added to the image point cloud data.

Next, in step S240, for each image point cloud data extracted in step S230, detailed image point cloud data is extracted using an algorithm capable of extracting the position of the edge of tooth 5a in units of sub-pixels. For extraction of detailed image point cloud data, for example, a brightness profile in a direction orthogonal to the contour of the gear 5 is obtained, and the position of the edge of the tooth 5a is extracted in units of subpixels by using the brightness profile. can be done. Here, it is possible to obtain normal line information perpendicular to the contour from the CAD data. Therefore, as described above, each comparison point cloud data has the normal line information.

Next, in step S250, using the detailed image point cloud data extracted in step S240, secondary matching is performed for each tooth 5a in the same manner as in step S230. In this step, since the result of the primary matching can be used, it is possible to omit performing matching while shifting the angle of the image point cloud data with respect to the comparison point cloud data by 30 degrees 12 times.

Next, in step S260, the comparison point cloud data matched in step S250 is compared with the detailed image point cloud data. Specifically, the error between the comparison point cloud data and the detailed image point cloud data is obtained. For example, if the image point cloud data has a longer distance from the center of gravity of the gear 5 than the image point cloud data, the comparison point cloud data has a longer distance from the center of gravity of the gear 5 than that of the comparison point cloud data. That is, in this step, the quality of the shape of each tooth 5a in the gear 5 is judged.

Next, in step S270, image point cloud data for two adjacent teeth extracted in step S220 or half teeth on both sides adjacent to one tooth extracted in step S220 are processed by steps similar to steps S230 to S250. Minute by minute image point cloud data is used to measure the pitch error between adjacent teeth 5a. That is, in this step, the quality of the pitch between the adjacent teeth 5a of the gear 5 is judged.

Then, in step S280, it is determined whether there is another gear 5 pattern-searched in step S130, and steps S140 to step S280 are repeated for all gears 5 pattern-searched in step S130, The component shape inspection method of the present embodiment is terminated. In this embodiment, as shown in FIG. 2, three gears 5 are pattern-searched in step S130, so steps S140 to S280 are repeated three times.

As described above, the component shape inspection method of the present embodiment corresponds to step S110, in which point cloud data, which is comparison data corresponding to a part of the contour of the gear 5, is read as comparison point cloud data. It has a group data reading step. Also, corresponding to step S120, there is an image data obtaining step of obtaining image data of the gear 5. FIG. Further, corresponding to step S140, there is a corresponding data obtaining step of obtaining corresponding data corresponding to the comparison data from the image data of the gear 5. FIG. Also, corresponding to steps S150, S210, S220 and S240, there is an image point cloud data extracting step of extracting image point cloud data, which is point cloud data, from the corresponding data. Also, corresponding to steps S260 and S270, there is a determination step of comparing the image point cloud data with the comparison point cloud data to determine whether the shape of the gear 5 is good or bad.

By executing the component shape inspection method of the present embodiment, the quality of the shape of the gear 5 is determined based on the range corresponding to a part of the contour of the gear 5, thereby inspecting the shape of the gear 5 with high accuracy. can do. This is because the outline of the gear 5 is a part that greatly affects the performance of the gear 5 and has a clear shape. In addition, instead of comparing the images of the entire gear 5, the images of a part of the outline of the gear 5 are compared, and the data of the point groups are compared. Therefore, it is possible to prevent the amount of data from becoming too large and reduce the inspection load. That is, the shape of the gear 5 can be easily inspected with high accuracy by executing the component shape inspection method of the present embodiment.

Further, the part that can be inspected by the part shape inspection method of the present embodiment is the gear 5, and the range corresponding to part of the contour of the gear 5 is the range that includes at least one tooth 5a of the gear 5. The gear 5 as a component is often used in a large amount in various devices, but by executing the component shape inspection method of the present embodiment, the shape of the gear 5 can be easily inspected with high accuracy. can.

Further, as described above, the comparison point cloud data is converted to have an accuracy of 0.1 μm, and is ten times or more the image point cloud data for measuring errors on the order of several μm. Therefore, by executing the component shape inspection method of the present embodiment, each image point cloud data can be compared with more preferable comparison point cloud data, and the shape of the gear 5 can be inspected with particularly high accuracy. can.

Further, as described above, the comparison point cloud data has normal line information, and in the comparison point cloud data reading step of step S110, the position of the contour portion of the gear 5 can be read based on the normal line information. Judging the data along the normal increases the accuracy of the data judgment. Since the position of the contour of the gear 5 is read, the data can be determined along the normal line, and the shape of the gear 5 can be inspected with particularly high accuracy.

Further, as described above, in the component shape inspection method of the present embodiment, as the image point cloud data extraction step, step S150 as the first image point cloud data extraction step for extracting image point cloud data in units of pixels; , and step S240 as a second image point cloud data extraction step for extracting image point cloud data in units of sub-pixels. That is, by executing the component shape inspection method of the present embodiment, rough inspection and detailed inspection can be performed, and the shape of the gear 5 can be inspected quickly and with high accuracy.

As described above, in the component shape inspection method of the present embodiment, data not used in the judgment steps of steps S260 and S270 are extracted from the image point cloud data extracted in the image point cloud data extraction step of step S150. It has an unused data extraction step corresponding to step S160. Therefore, by executing the component shape inspection method of this embodiment, the shape of the gear 5 can be inspected with particularly high accuracy.

As described above, in the component shape inspection method of the present embodiment, the data conversion step of converting the image point cloud data extracted in the image point cloud data extraction step of step S150 corresponds to steps S170 and S190. have. In the determination process of steps S260 and S270, the quality of the shape of the gear 5 can be determined using the data converted in the data conversion process. Therefore, by executing the part shape inspection method of the present embodiment and, for example, executing step S170, a circular image is converted into a belt-shaped image so that a plurality of similar shape portions can be compared quickly and with high accuracy. can be done. Further, by executing the component shape inspection method of the present embodiment and, for example, executing step S190, it becomes possible to correct the center position of the image data acquired with the deviation of the center position. The shape of the part can be inspected.

It goes without saying that the present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the invention described in the claims, and these modifications are also included in the scope of the present invention.

1... Parts shape inspection system, 2... PC, 2a... Monitor,
2b...storage unit, 3...camera, 4...telecentric lens, 5...gear (component),
5a...teeth, 5c...center, 6...stage, 7...illumination device, P1...start point, P2...end point

Claims (6)

A method for inspecting the shape of a component,
a comparison point cloud data reading step of reading point cloud data, which is comparison data corresponding to a part of the outline of the part, as comparison point cloud data;
an image data acquisition step of acquiring image data of the component;
a corresponding data acquisition step of acquiring corresponding data corresponding to the comparison data from the image data of the component;
an image point cloud data extracting step of extracting image point cloud data, which is point cloud data, from the corresponding data;
a determination step of comparing the image point cloud data with the comparison point cloud data to determine whether the shape of the component is good or bad ;
an unused data extraction step of extracting data not used in the determination step from the image point cloud data extracted in the image point cloud data extraction step;
has
In the unused data extracting step, the image point cloud data is labeled and grouped, and data out of the image point cloud data in a group having the largest area among the plurality of grouped groups is regarded as unused data. A method for inspecting the shape of a part, characterized by: In the component shape inspection method according to claim 1,
A method for inspecting the shape of a component, wherein the component is a gear, and the range corresponding to a portion of the contour of the component includes at least one tooth of the gear. In the component shape inspection method according to claim 1 or 2,
A method for inspecting the shape of a part, wherein the number of comparison point cloud data is ten times or more that of the image point cloud data. In the component shape inspection method according to any one of claims 1 to 3,
The comparison point cloud data has normal line information,
The component shape inspection method, wherein the comparison point cloud data reading step reads position data of a contour portion of the component based on the normal line information. In the component shape inspection method according to any one of claims 1 to 4,
The image point cloud data extraction step includes a first image point cloud data extraction step of extracting the image point cloud data in pixel units and a second image point cloud data extraction step of extracting the image point cloud data in subpixel units. A method for inspecting the shape of a part, comprising: In the component shape inspection method according to any one of claims 1 to 5 ,
a data conversion step of converting the image point cloud data extracted in the image point cloud data extraction step;
A method for inspecting a shape of a component, wherein the determination step uses the data converted in the data conversion step to determine whether the shape of the component is good or bad.

Images