Machining quality inspection method and system for watch bezel of manipulator watch
By analyzing multiple images of the watch bezel, calculating the normal vibration distance range and pixel point anomalies, the problem of unqualified bezels being affected by reflection and jitter during detection is solved, and more accurate defect detection is achieved.
Patent Information
- Application Number
- CN202510421115.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-07
- Publication Date
- 2025-05-06
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
When using PVD technology to coat the watch bezel, the unqualified bezel is easily affected by the bezel reflective area and jitter during the inspection process, resulting in inaccurate defect detection results.
By acquiring multiple bezel images, analyzing the normal and abnormal areas of each bezel, calculating the range of normal vibration distances of each bezel, and determining the final anomaly of each pixel point based on the difference in grayscale values of the bezel, so as to accurately identify the bezel with unqualified coatings.
The impact of the bezel reflective area on the defect detection results is effectively reduced, and the impact of jitter on the detection results is reduced by identifying the normal jitter range, thereby improving the accurate detection rate of the unqualified bezel of coating.
Smart Images
Figure CN119941718A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of image processing, and in particular to a processing quality inspection method and system for a mechanical watch bezel. Background Art
[0002] The watch bezel is an important part of the watch. It not only protects the watch dial, but also beautifies the appearance of the watch. During the production process of the watch bezel, the bezel is first polished to make the appearance of the bezel meet the specified standards. Then, in order to increase the service life of the watch bezel and increase the wear resistance of the bezel, the bezel is PVD coated. The coated bezel is sent to the watch assembly factory for installation.
[0003] When using PVD technology to coat bezels, a large number of bezels are coated at the same time. After each coating, there may be unqualified bezels. It is necessary to perform abnormal detection on the coated bezels to screen out the unqualified bezels. When comparing each bezel with the surrounding bezels to obtain the unqualified bezels, because the bezels are in a vacuum environment when the PVD technology is used to coat the bezels, and when the bezels are photographed, the vacuum environment of the bezels changes, causing the bezels on the electroplating rack to shake, and the shaking will affect the screening results. Summary of the invention
[0004] The present invention provides a processing quality inspection method and system for a mechanical watch bezel, so as to solve the problem that the reflective area on the bezel will interfere with the defect detection result when the defect detection is performed on the electroplated bezel.
[0005] The present invention provides a mechanical watch bezel processing quality inspection method and system using the following technical solutions: The present invention provides a processing quality inspection method for a mechanical watch bezel, the method comprising the following steps: Acquire multiple bezel images from each camera angle under each illumination, and obtain the bezel in each bezel image; Obtaining a normal area and an abnormal area of each bezel in each bezel image; obtaining a range of a normal vibration distance of each bezel in the bezel image at each camera angle under each lighting condition based on changes in the normal area and the abnormal area in multiple bezel images of the same bezel at the same camera angle under each lighting condition; Obtain multiple corresponding bezels and multiple reference bezels of the same layer for each bezel under each light source; obtain the final abnormality of the pixel points of each bezel based on the difference between the grayscale values of the corresponding bezels and the pixel points at the same position in the reference bezels whose vibration distances are all within the range of the normal vibration distance of the same bezel, and then obtain the bezel with unqualified coating.
[0006] Furthermore, the specific method of obtaining the normal area and the abnormal area of each bezel in each bezel image includes: Perform edge detection on a bezel in any bezel image to obtain multiple detection areas on each bezel; and obtain a normal area and an abnormal area of each bezel according to the area distribution of the detection areas on each bezel.
[0007] Furthermore, the specific method of obtaining the range of normal vibration distance of each bezel in the bezel image at each camera angle under each illumination includes: The first Under the light The vertical coordinate of all bezel images from the camera angle is The image in the bezel The vertical coordinate of the center of mass of the first bezel is the same as that of the first The mean of the horizontal coordinates of the centroids of the first bezel is taken as The horizontal coordinate of the center of mass of the fixed bezel of each bezel is obtained to obtain the vibration distance of each bezel in each bezel image and the degree of variation of the defect characteristics of each bezel; According to the vibration distance of each bezel in each bezel image and the variation degree of the defect characteristic of each bezel, the range of the normal vibration distance of each bezel at each camera angle under each lighting is obtained.
[0008] Furthermore, the specific method of obtaining the vibration distance of each bezel in each bezel image and the defect characteristic variation degree of each bezel is as follows: The first Under the light Camera angle The image in the bezel The Euclidean distance between the center of mass of the first bezel and its fixed bezel is denoted as The vibration distance of each bezel; According to the said The Euclidean distance between the centroid of the defect area of the first bezel and its fixed bezel is used as the influence weight of the defect area on the defect feature variation degree. The Euclidean distance between the center of mass of the defective area of the first bezel and the fixed bezel and the center of mass of the normal area is obtained. The degree of variation of defect characteristics of each bezel.
[0009] Furthermore, the specific method of obtaining the range of normal vibration distance of each bezel at each camera angle under each lighting condition includes: The coordinate system is constructed with the vibration distance as the horizontal axis and the defect characteristic variation as the vertical axis. Under the light The centroid of all bezel images from the camera angles is The center of mass of the fixed bezel of the first bezel is on the left and on the right Each circle is mapped separately to obtain multiple data points; The number of data points to be fitted is increased from left to right with a step size of one, and the data points in the two-dimensional coordinate system are gradually fitted; according to the change of the fitting error of each data point in the multiple fitting processes, the first Under the light The bezel image from each camera angle The range of normal vibration distance of the bezel.
[0010] Furthermore, the specific method of obtaining a plurality of corresponding bezels of each bezel under each illumination and a plurality of comparison bezels of the same layer includes: The first In all bezel images at each camera angle under each lighting The maximum value of the average area of the bezels corresponds to the first bezel, recorded as Under the light A number of corresponding bezels for a bezel; According to each bezel image and the The difference in the vertical coordinates of the center of mass of the first bezel is obtained The same layer of bezels; According to In all bezel images at each camera angle under each lighting The first The average area of the same layer of the bezel is Under the light The degree of proximity of the area averages of all corresponding bezels of the bezel is obtained. The first Several comparison circles of the same layer circle.
[0011] Furthermore, the specific method of obtaining the final abnormality of the pixel points of each bezel is as follows: Perform edge detection on the bezel to obtain a number of detection areas of each bezel, and obtain a reflective area and a non-reflective area of each bezel according to an average gray value of pixels in the detection area of each bezel; Jordi Under the light The first The corresponding bezel and The first The first If the vibration distances of the first and second comparison bezels are all within the range of their normal vibration distances, the grayscale values of the pixels located in the reflective area, the non-reflective area, and the same position in different areas of the two bezels are fitted respectively, and the grayscale values of the pixels located in the reflective area, the non-reflective area, and the same position in different areas of the two bezels are obtained. A horizontal initial abnormality of the pixel points of the bezel; The product of the defect characteristic variation of the two bezels is taken as the initial lateral abnormality for the first The pixels of the bezel are The influence weight of the lateral anomaly under the illumination is obtained as The first The pixel at The lateral anomaly under each light is calculated to obtain the final anomaly of each pixel point in each bezel.
[0012] Furthermore, the specific method of obtaining the final abnormality of each pixel point of each bezel is as follows: According to the difference between the abscissa of the centroid of each bezel and other bezels in each bezel image, the same column of each bezel is obtained; The first The pixel point at the same position on the circle and all the circles with the same column is The sum of the absolute values of the differences of the lateral anomaly under the first illumination and the The first The pixel at The product of the lateral anomaly under the illumination is taken as the The first The pixel at The degree of abnormal performance under each light; According to the abnormal performance degree of each pixel point of each bezel under each illumination, the final abnormality degree of each pixel point of each bezel is obtained.
[0013] Furthermore, the specific method of obtaining the final abnormality of each pixel point of each bezel is as follows: All pixels of a bezel are The average value of the lateral anomaly under the illumination is taken as the value of the pixel point on the bezel at the The influence weight of the abnormal expression degree under each lighting on the final abnormality degree of the pixel point is calculated to obtain the final abnormality degree of each pixel point in each bezel.
[0014] The present invention also proposes a processing quality inspection system for a mechanical watch bezel, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the above method.
[0015] The beneficial effects of the technical solution of the present invention are as follows: when obtaining an unqualified bezel with electroplating on a plating rack, due to the reflective area on the surface of the bezel, the normal pixel points in the reflective area of one bezel are greatly different from the normal pixel points in the non-reflective area, and the defective pixel points cannot be accurately obtained by only one bezel. The present method uses the qualified bezels to be similar in all aspects, and the bezels that have been electroplated all have the characteristics of the reflective area. According to the grayscale difference of the pixel points at the same position on each bezel that has been electroplated and on other bezels that have been electroplated and are both in the reflective area or not in the reflective area, the possibility of each pixel point of the bezel being a defective pixel point is calculated, thereby reducing the influence of the reflective area on the defect detection result; by comparing different bezels, since the bezel is in a vacuum during the PVD electroplating process, the bezel In an environment, when taking a photo of the plating rack containing the bezel, the vacuum environment of the bezel will be destroyed, causing the bezel to shake when taking the photo, and affecting the defect detection result; when the bezel shakes, the relationship between the pixels on the bezel and the light changes, which makes the jitter affect the edge detection result of the bezel, according to the jitter range where the jitter has a smaller impact on the defect detection result, the change of the jitter distance of the bezel and the edge detection result of the bezel, and the jitter range where the jitter has a greater impact on the defect detection result, the change of the jitter distance of the bezel and the edge detection result of the bezel are greatly different, the normal jitter range of each bezel is obtained, the impact of the jitter on the defect detection result is reduced, and the defective pixel points on each bezel are acquired more accurately. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0017] Figure 1 A flowchart of the steps of a processing quality inspection method for a mechanical watch bezel according to the present invention; Figure 2 This is the production process flow chart of watches. DETAILED DESCRIPTION
[0018] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the processing quality inspection method and system of a mechanical watch bezel proposed by the present invention, its specific implementation method, structure, features and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.
[0019] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
[0020] The specific scheme of the processing quality inspection method and system for a mechanical watch bezel provided by the present invention is described in detail below with reference to the accompanying drawings.
[0021] See also Figure 1 , which shows a flowchart of a method for processing and inspecting a mechanical watch bezel according to an embodiment of the present invention, the method comprising the following steps: Step S001: Acquire multiple bezel images from each camera angle under each lighting, and obtain the bezel in each bezel image.
[0022] Specifically, a camera is installed above the electroplating rack and rotates around the electroplating rack to take pictures of the electroplating rack, and multiple images of the electroplating rack with the electroplated bezels are obtained under multiple lighting conditions and at different camera angles. When taking pictures of the electroplating rack, multiple images of the electroplating rack with the electroplated bezels are taken at each camera angle under each lighting condition. , the number of preset camera angles , this is used as an example for description, and other values may be set in other implementation modes.
[0023] Furthermore, all circles in each circle image are obtained by semantic segmentation. Semantic segmentation is a well-known technology in neural networks, that is, circles in a large number of circle images are labeled, and a semantic segmentation model is constructed by training a neural network, and a cross entropy loss function is used as the loss function. Figure 2 This is the production process flow chart of watches.
[0024] Thus, a plurality of bezel images are obtained, and a bezel in each bezel image is obtained.
[0025] Step S002: Obtain the normal area and abnormal area of each bezel in each bezel image; obtain the range of normal vibration distance of each bezel in the bezel image at each camera angle under each lighting according to the changes in the normal area and abnormal area in multiple bezel images of the same bezel at the same camera angle under each lighting.
[0026] It should be noted that when taking a photo of a coated bezel, since the coated bezel is relatively smooth, it contains a reflective area, and the reflective area is significantly different from the normal area. When the edge detection algorithm is used to obtain the abnormal area on the coated bezel, the reflective area will be regarded as the abnormal area, resulting in the existing edge detection algorithm being unable to directly obtain the abnormal area on the coated bezel. Therefore, the existing method is improved.
[0027] It should be further explained that the two qualified bezels after electroplating have strong consistency in all aspects, that is, under the same light, the grayscale values of the pixels at the same position on the two qualified bezels after electroplating are relatively similar. Therefore, by comparing the grayscale of the pixels on each bezel with the pixels at the same position on the surrounding bezels, the defective pixels on each bezel are obtained.
[0028] It should be further explained that when the bezel is coated with PVD technology, the bezel is coated in a vacuum environment. When taking pictures of the coating frame, a camera will be installed above the coating frame to change the vacuum environment of the bezel into a non-vacuum environment, resulting in part of the bezel on the coating frame shaking due to the change in the environment when taking pictures of the bezel.
[0029] It should be further explained that the grayscale value of each pixel is related to the light received by the pixel, and the light is related to the distance between the pixel and the light source. When the bezel is shaken, the distance between different pixels and the light source and the angle relationship between the cameras change differently, so that when the bezel is shaken, the difference between normal pixels and abnormal pixels will change, resulting in the jitter affecting the defect detection results when defective pixels are obtained by comparing different bezels. Therefore, the jitter range that has a smaller impact on the defect detection results is obtained, that is, the range of normal vibration distance of each bezel.
[0030] It should be further explained that, when the bezel is shaken, the difference between normal pixels and abnormal pixels will change, and the edge detection result on the bezel will also change. Therefore, according to the change of the edge detection result when the bezel is shaken, the influence of the shake on the edge detection result is obtained, and it is obtained that when the bezel is shaken within a certain range, the shake has a greater influence on the edge detection result, and when the bezel is shaken within a certain range, the shake has a smaller influence on the edge detection result.
[0031] It should be further explained that when obtaining the change of edge detection results during the shaking process, it is actually to observe the change of normal area and abnormal area when the bezel is shaking. Since the number of defective pixels on each bezel is small, the largest detection area is taken as the normal area and the other areas are taken as abnormal areas in the edge detection result of each bezel. According to the change of normal area and abnormal area in multiple bezel images of each bezel during the shaking process, the range of normal vibration distance of the bezel is obtained.
[0032] It should be further explained that if the vibration of the bezel has a small impact on the normal area and the abnormal area when the bezel vibrates within a range, and has a large impact on the normal area and the abnormal area when the bezel vibrates within another range, then the changes in the normal area and the abnormal area of the bezel within these two vibration ranges are greatly different from the fitting formula of the vibration distance. Therefore, the fitting method is used to obtain the range of the normal vibration distance of the bezel.
[0033] It should be further explained that when obtaining the jitter distance of the bezel, it is necessary to first obtain the position of the bezel when it is not jittering. Since the bezel can be regarded as a symmetrical movement when it is jittering, the position of the bezel when it is not jittering is obtained based on the position of the bezel when it is jittered multiple times, and the jitter distance of each bezel is obtained.
[0034] It should be further explained that when obtaining the change of the normal area and the abnormal area of the bezel, when the normal area and the abnormal area of the bezel change, the centroid of the normal area and the abnormal area will also change. Therefore, according to the change of the centroid of the normal area and the abnormal area when the bezel shakes once, the degree of change of the defect characteristic of the bezel during this shaking process is obtained.
[0035] Specifically, for Under the light Camera angle The image in the bezel The bezel is edge detected to obtain multiple detection areas on the bezel. The detection area with the largest area on the bezel is recorded as the normal area of the bezel, and the other areas are recorded as the abnormal areas of the bezel. Among them, edge detection of an image is a well-known technology, which will not be described in detail in this embodiment; it should be noted that the first detection area in all bezel images is not limited by lighting and camera angle. Each bezel corresponds to the same bezel.
[0036] Further, get the Under the light Camera angle The image in the bezel The horizontal coordinate and vertical coordinate of the centroid of the bezel. Obtaining the centroid of a graphic is a well-known technique and will not be described in detail in this embodiment. Under the light All bezel images from all camera angles are consistent with the The image in the bezel The first bezel, recorded as Under the light Camera angle The image in the bezel Symmetrical bezel with two bezels.
[0037] Further, get the Under the light Camera angle The specific calculation formula for the horizontal coordinate of the center of mass when the bezel does not vibrate is as follows: In the formula, Indicates Under the light Camera angle The horizontal coordinate of the center of mass when the bezel does not vibrate, Indicates Under the light Camera angle The image in the bezel The horizontal coordinate of the centroid of the bezel, Indicates Under the light Camera angle The image in the bezel The mean of the abscissas of the centroids of all symmetric circles of a circle.
[0038] It should be noted that since the bezel is in a symmetrical motion when it is shaking, the average value of the horizontal coordinates of the two bezels with the same vertical coordinates when the bezel is shaking is , which is the horizontal coordinate of the center of mass when the bezel is not shaking.
[0039] Furthermore, Under the light The horizontal coordinates of the centroids in all the circle images of the camera angles are When the bezel is not shaking, the horizontal coordinate of the center of mass is the same as the bezel, recorded as Under the light Camera angle Several fixed bezels with a bezel.
[0040] Furthermore, Under the light Camera angle The center of mass of the fixed bezel of the first bezel is Under the light Camera angle The image in the bezel The Euclidean distance between the centroids of the first bezel is Under the light Camera angle The image in the bezel The vibration distance of the bezel.
[0041] Further, get the Under the light Camera angle The image in the bezel The centroid of the normal area and the centroid of the abnormal area on the bezel are obtained. Under the light Camera angle The image in the bezel The specific calculation formula for the defect characteristic variation of each bezel is as follows: In the formula, Indicates Under the light Camera angle The image in the bezel The variation of defect characteristics of each bezel, Indicates Under the light Camera angle The image in the bezel The first bezel and the The Euclidean distance between the centroids of the normal areas on the fixed bezel of the bezel, Indicates Under the light Camera angle The image in the bezel The first bezel and the The Euclidean distance between the centroids of the anomaly areas on the fixed bezel of the bezel, express Function, this embodiment is used for normalization processing.
[0042] It should be noted that when When the abnormal area of the bezel changes greatly during vibration, The value of is large. As , to increase the The influence of the change of abnormal area on the variation of defect characteristics when the first bezel shakes; When the normal area of the bezel changes greatly during vibration, the abnormal area The value of is large. As , to increase the The influence of the change of abnormal area on the variation of defect characteristics when the bezel vibrates.
[0043] Furthermore, a two-dimensional coordinate system is constructed with the vibration distance as the horizontal axis and the defect characteristic variation as the vertical axis. Under the light The horizontal coordinate of the centroid of all bezel images from the camera angle is less than or equal to the The horizontal coordinate of the center of mass of the fixed bezel is The first bezel is marked as The left mapping circles of the circle, for the A plurality of left mapping circles of a circle are mapped to obtain a plurality of data points.
[0044] Furthermore, the least squares method is used to gradually fit the data points in the two-dimensional coordinate system from left to right. The specific fitting process is as follows: first, the least squares method is used to fit the two leftmost data points to obtain the first fitting error of the two leftmost data points, and this fitting process is recorded as the first fitting process; the least squares method is used to fit the three leftmost data points to obtain the fitting error of the first three leftmost data points in the fitting process, and this fitting process is recorded as the second fitting process; the least squares method is used to fit the four leftmost data points to obtain the fitting error of the first four leftmost data points in the fitting process, and this fitting process is recorded as the third fitting process; the number of data points to be fitted is continuously increased from left to right with a step size of one.
[0045] Furthermore, in the process of fitting the data points in the two-dimensional coordinate system, if The data point in The fitting error in the first fitting process is smaller than that in the second fitting process. The fitting error in the first fitting process is The data point in The fitting error in the first fitting process is greater than that in the second fitting process. The fitting error in the first fitting process is The vibration distance of the data point is Under the light All bezel images from all camera angles The maximum normal vibration distance of the left mapping bezel of the bezel; among them, , ; The quantity threshold value preset in this embodiment , this is used as an example for description, and other values may be set in other implementation modes.
[0046] It should be noted that when fitting the data points in the coordinate system, if The fitting formula obtained by fitting the data points to the left of the data point is the same as that of the When the fitting formula obtained after fitting the data points to the right of the data point is quite different, as the number of The data point to the right of the data point The fitting error of the data point is gradually decreasing during the fitting process; The fitting errors of the data points on the left side of the data point gradually increase during the fitting process.
[0047] Similarly, get Under the light The horizontal coordinate of the centroid of all bezel images from the camera angle is greater than The horizontal coordinate of the center of mass of the fixed bezel is bezel, and record it as The maximum normal vibration distance of the right mapping bezel is obtained by mapping the right mapping bezel to the right of the bezel. Under the light The bezel image from each camera angle The range of normal vibration distance of the bezel.
[0048] Step S003: Obtain multiple corresponding bezels and multiple reference bezels of the same layer for each bezel under each light source; obtain the final abnormality of the pixel points of each bezel based on the difference between the grayscale values of the corresponding bezels and the pixel points at the same position in the reference bezels whose vibration distances are all within the range of the normal vibration distance of the same bezel, and then obtain the bezel with unqualified coating.
[0049] It should be noted that when comparing two bezels within the normal vibration distance to obtain the lateral abnormality of each bezel pixel, the larger the bezel in the monitoring image, the more information the bezel conveys. When comparing a bezel to its surrounding bezels, select The surveillance image under the camera angle corresponding to the largest bezel area Each bezel is compared with its surrounding bezels to obtain the lateral abnormality of each bezel pixel.
[0050] It should be further explained that when a camera is used to shoot the bezels on the electroplating rack, the angles and distances between different bezels on the same layer of the electroplating rack and the camera are different, so that the sizes of different bezels on the same layer in the surveillance image shot at a camera angle may be quite different. When comparing a bezel in an image with its surrounding bezels, a bezel needs to be scaled to a large extent, resulting in low accuracy of the lateral abnormality of each pixel. When a bezel is compared with its surrounding bezels, the first bezel in the surveillance image from a camera angle under the same lighting can be The first bezel in the surveillance image from other camera angles Compare the bezels around the bezel to get The lateral anomaly of each pixel on the bezel.
[0051] It should be further explained that the first The bezel and other monitoring images When comparing the bezel around the first bezel, choose Compare with the bezel with the closest bezel area to get the The lateral anomaly of each pixel on the bezel.
[0052] It should be further explained that, since the electroplating rack is photographed under multiple lighting conditions, the lateral abnormality of each bezel pixel point is obtained according to the lateral initial abnormality of each bezel pixel point under multiple lighting conditions.
[0053] It should be further explained that when obtaining defective pixels based on the lateral abnormality of each pixel, a threshold is set, and defective pixels with lateral abnormality greater than the threshold are regarded as defective pixels, and other pixels are regarded as normal pixels. Since the difference between normal pixels and defective pixels in the reflective area is much smaller than the difference between normal pixels and defective pixels in the non-reflective area, although the lateral abnormality of the defective pixels in the reflective area is somewhat different from that of the normal pixels, the difference is not obvious. Therefore, the difference in abnormality between defective pixels and normal pixels is increased.
[0054] It should be further explained that, since the electroplating rack is a multi-layer structure, the lateral abnormality of the pixel points of each bezel is corrected by comparing each bezel with other bezels in the same column. Since the qualified bezels are the majority when the bezels are electroplated, that is, most of the pixels on each electroplated bezel are normal, and there is a certain difference in the lateral abnormality between normal pixels and defective pixels, the lateral abnormality of each pixel point is corrected by subtracting the lateral abnormality of the pixel point on each bezel from the pixel point at the same position in the same column, and the abnormal performance of each pixel point on each bezel under each light is obtained.
[0055] It should be further explained that, since the coating rack is a relatively high rack, and the degree to which each bezel is affected by light is negatively correlated with the distance between the bezel and the light source, when the upper bezel of the coating rack is overexposed, that is, the difference between normal pixels and abnormal pixels is small, the lower bezel may receive a more appropriate light intensity, that is, there is a more obvious difference between normal pixels and abnormal pixels; and when the light intensity received by the upper bezel of the coating rack is more appropriate, the light intensity received by the lower bezel may be insufficient, so different weights are assigned to the lateral abnormality under different illumination when calculating the final abnormality of each pixel. Since the lateral abnormality of all pixels on a bezel under each illumination reflects the grayscale difference between normal pixels and abnormal pixels of the bezel under this illumination, the sum of the lateral abnormality of all pixels on a bezel under each illumination is used as the weight of the abnormal performance of the pixel under this illumination on the final abnormality.
[0056] Specifically, the In all bezel images at each camera angle under each lighting The camera angle corresponding to the maximum value of the average area of the bezel is recorded as In the image of the bezel under the lighting The comparison camera angle of the first bezel. In the image of the bezel under the lighting Under the comparison camera angle of the bezel, the first bezel, recorded as Under the light Several corresponding bezels for a bezel.
[0057] Furthermore, the centroid ordinate in each bezel image is compared with the The absolute value of the difference between the vertical coordinates of the centroids of the bezels is within the vertical axis difference threshold The bezel below is marked as The vertical axis difference threshold preset in this embodiment is , this is used as an example for description, and other values may be set in other implementation modes.
[0058] Furthermore, In all bezel images at each camera angle under each lighting The first The average area of the same layer of the table circle, and the Under the light The camera angle corresponding to the area of all corresponding bezels of the bezel is closest to the average area of the bezel, which is recorded as Under the light The first The comparison camera angle of the same layer circle. Under the light The first The comparison of the same layer bezels in the camera angle of each bezel image The first The same layer of the bezel is recorded as The first Several comparison circles of the same layer circle.
[0059] Further, edge detection is performed on any bezel to obtain a number of edges, which are surrounded by a number of detection areas, and then a number of detection areas of the bezel are obtained. The detection area in which the average grayscale value of the pixel points in the detection area is greater than the average grayscale value of all the pixel points on the bezel is recorded as the reflective area of the bezel; the other detection areas are recorded as the non-reflective areas of the bezel. The corresponding bezel of the first bezel is The reference bezel of the same layer of the first bezel is scaled so that the The corresponding bezel of the first bezel is The areas of several comparison circles of the same layer of the circle are the same. The image pyramid algorithm is a well-known technology, which will not be described in detail in this embodiment.
[0060] Further, the Under the light The grayscale value of the pixel point of the corresponding bezel is the horizontal axis, The first The grayscale value of the pixel points of the comparison circle of the same layer is used as the vertical axis to construct an updated coordinate system.
[0061] Furthermore, if Under the light The first The corresponding bezel and The first The first The vibration distance of the reference bezel is within the normal vibration distance range. The first The corresponding bezel and The first The first The grayscale values of the pixels at the same position in the reflective area on the reference table ring are mapped to the updated coordinate system to obtain a number of sample points in the updated coordinate system; the sample points in the updated coordinate system are fitted using the least squares method to obtain the fitting error of each sample point. The least squares method is a well-known technology and will not be described in detail in this embodiment.
[0062] Furthermore, according to The first The corresponding bezel and The first The first The grayscale values of the pixels at the same position in the non-reflective area on the reference table ring are mapped to the updated coordinate system to obtain a number of coordinate points in the updated coordinate system; the coordinate points in the updated coordinate system are fitted using the least squares method to obtain the fitting error of each coordinate point.
[0063] Further, in the The first The corresponding bezel and The first The first The pixel points at the same position of the reference table circle and not mapped to the updated coordinate system are mapped to the updated coordinate system to obtain several observation points in the updated coordinate system; the observation points in the updated coordinate system are fitted using the least squares method to obtain the fitting error of each observation point.
[0064] Furthermore, the fitting error of each sample point is recorded as The horizontal initial abnormality of the pixel point of the bezel is recorded as the fitting error of each coordinate point corresponding to the first The horizontal initial abnormality of the pixel point of the circle is calculated; the fitting error of each observation point is recorded as the first The horizontal initial abnormality of the pixel points of the bezel is The first The corresponding bezel and The first The first The product of the defect characteristic variation degrees of the reference table circles is recorded as the jitter influence degree of the corresponding lateral initial abnormality; it should be noted that the lateral initial abnormalities of several sample points obtained under a group of corresponding table circles and reference table circles all correspond to the jitter influence degrees obtained under the group of corresponding table circles and reference table circles; any pixel point on a table circle corresponds to a lateral initial abnormality under multiple groups of corresponding table circles and reference table circles.
[0065] Further, get the The first The pixel at The specific calculation formula of the lateral anomaly under illumination is as follows: In the formula, Indicates The first The pixel at The lateral anomaly under each illumination is Indicates The first The pixel at The number of lateral initial anomalies under illumination, Indicates The first The pixel at The first The initial lateral anomaly, Indicates The first The pixel at The first The jitter influence of the initial lateral anomaly, represents the weight normalization function, To prevent the hyperparameter from having a denominator of 0, this embodiment sets , this is used as an example for description, and other values may be set in other implementation modes.
[0066] It should be noted that The larger the value of The first The corresponding bezel and The first The first The greater the impact of the vibration on the reference bezel, the The first The pixel at The first The greater the impact of jitter on the initial lateral anomaly, the less reliable it is. As The weight of .
[0067] Furthermore, the horizontal coordinate of the centroid in each bezel image is compared with the The absolute value of the difference between the horizontal coordinates of the centroids of the bezels is less than the horizontal axis difference threshold The bezel is recorded as The horizontal axis difference threshold preset in this embodiment is , this is used as an example for description, and other values may be set in other implementation modes.
[0068] Furthermore, the image pyramid algorithm is used to The first circle and its corresponding circle are scaled so that The area of each circle is the same as that of the circle in the same column. The image pyramid algorithm is a well-known technology and will not be described in detail in this embodiment.
[0069] Furthermore, according to the lateral abnormality of each bezel pixel point, the specific calculation formula for the abnormal performance of each bezel pixel point under each illumination is obtained as follows: In the formula, Indicates The first The pixel at The abnormal performance under light, Indicates The first The pixel at The lateral anomaly under each illumination is Indicates The first The first The pixel at The lateral anomaly under each illumination is represents the absolute value function, Indicates the number of circles in the same table for each table circle.
[0070] It should be noted that The larger the value, the The first The greater the difference in lateral abnormality between the pixel point and the pixel points at the same position on the same row of other bezels, the more it conforms to the characteristic that the lateral abnormality of normal pixels and defective pixels is different, indicating that the The first The greater the possibility that a pixel is a defective pixel.
[0071] Furthermore, the specific calculation formula for obtaining the final abnormality of each pixel point in each bezel is as follows: In the formula, Indicates The first The final abnormality of the pixel point, Indicates the number of lights. Indicates All pixels of the bezel are in The mean of the lateral anomaly under illumination, Indicates The first The pixel at The abnormal performance under light, represents the weight normalization function, Represents the weight normalization function, whose input object is the The sum of the lateral anomalies of all pixels in a circle under each illumination; express Function, this embodiment is used for normalization processing.
[0072] It should be noted that The larger the value, the Under the light The greater the difference between the defective pixels and the normal pixels on the first bezel, the The first The pixel at The more credible the abnormal performance under each light, the more reliable the abnormal performance under each light. All pixels of the bezel are in The sum of the lateral anomalies under illumination is obtained as The first The pixel at The abnormal performance of the first The first The impact of the final abnormality of each pixel.
[0073] At this point, the final abnormality of each pixel point in each bezel is obtained.
[0074] Furthermore, a preset abnormality threshold , if the final abnormality of a pixel in a circle is greater than the abnormality threshold , then the pixel of the bezel is a defective pixel. If the ratio of the defective pixel of a bezel to all pixels is greater than the defect level threshold , then the bezel is an unqualified bezel with coating. Among them, the abnormality threshold preset in this embodiment is , defect level threshold , this is used as an example for description, and other values may be set in other implementation modes.
[0075] At this point, this embodiment is completed.
[0076] Another embodiment of the present invention provides a processing quality inspection system for a mechanical watch bezel, the system comprising a memory, a processor, and a computer program stored in the memory and running on the processor, and when the processor executes the computer program, the above method steps S001 to S003 are implemented.
[0077] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A processing quality inspection method for a mechanical watch bezel, characterized in that: The method comprises the following steps: Acquire multiple bezel images from each camera angle under each illumination, and obtain the bezel in each bezel image; Obtaining a normal area and an abnormal area of each bezel in each bezel image; obtaining a range of a normal vibration distance of each bezel in the bezel image at each camera angle under each lighting condition based on changes in the normal area and the abnormal area in multiple bezel images of the same bezel at the same camera angle under each lighting condition; Obtain multiple corresponding bezels and multiple reference bezels of the same layer for each bezel under each light source; obtain the final abnormality of the pixel points of each bezel based on the difference between the grayscale values of the corresponding bezels and the pixel points at the same position in the reference bezels whose vibration distances are all within the range of the normal vibration distance of the same bezel, and then obtain the bezel with unqualified coating.
2. A method for processing and inspecting the bezel of a mechanical watch according to claim 1, characterized in that: The specific method of obtaining the normal area and the abnormal area of each bezel in each bezel image includes: Perform edge detection on a bezel in any bezel image to obtain multiple detection areas on each bezel; and obtain a normal area and an abnormal area of each bezel according to the area distribution of the detection areas on each bezel.
3. The method for processing and inspecting the bezel of a mechanical watch according to claim 1, characterized in that: The specific method of obtaining the range of normal vibration distance of each bezel in the bezel image at each camera angle under each lighting condition includes: The first Under the light The vertical coordinate of all bezel images from the camera angle is The image in the bezel The vertical coordinate of the center of mass of the first bezel is the same as that of the first The mean of the horizontal coordinates of the centroids of the first bezel is taken as The horizontal coordinate of the center of mass of the fixed bezel of each bezel is obtained to obtain the vibration distance of each bezel in each bezel image and the degree of variation of the defect characteristics of each bezel; According to the vibration distance of each bezel in each bezel image and the variation degree of the defect characteristic of each bezel, the range of the normal vibration distance of each bezel at each camera angle under each lighting is obtained.
4. The method for processing and inspecting the bezel of a mechanical watch according to claim 3, characterized in that: The specific method of obtaining the vibration distance of each bezel in each bezel image and the defect characteristic variation degree of each bezel is as follows: The first Under the light Camera angle The image in the bezel The Euclidean distance between the center of mass of the first bezel and its fixed bezel is denoted as The vibration distance of each bezel; According to the said The Euclidean distance between the centroid of the defect area of the first bezel and its fixed bezel is used as the influence weight of the defect area on the defect feature variation degree. The Euclidean distance between the center of mass of the defective area of the first bezel and the fixed bezel and the center of mass of the normal area is obtained. The degree of variation of defect characteristics of each bezel.
5. The method for processing and inspecting the bezel of a mechanical watch according to claim 3, characterized in that: The specific method of obtaining the range of normal vibration distance of each bezel at each camera angle under each lighting condition includes: The coordinate system is constructed with the vibration distance as the horizontal axis and the defect characteristic variation as the vertical axis. Under the light The centroid of all bezel images from the camera angles is The center of mass of the fixed bezel of the first bezel is on the left and on the right Each circle is mapped separately to obtain multiple data points; The number of data points to be fitted is increased from left to right with a step size of one, and the data points in the two-dimensional coordinate system are gradually fitted; according to the change of the fitting error of each data point in the multiple fitting processes, the first Under the light The bezel image from each camera angle The range of normal vibration distance of the bezel.
6. The method for processing and inspecting the bezel of a mechanical watch according to claim 1, characterized in that: The specific method of obtaining a plurality of corresponding bezels of each bezel under each illumination and a plurality of comparison bezels of the same layer includes: The first In all bezel images at each camera angle under each lighting The maximum value of the average area of the bezels corresponds to the first bezel, recorded as Under the light A number of corresponding bezels for a bezel; According to each bezel image and the The difference in the vertical coordinates of the center of mass of the first bezel is obtained The same layer of bezels; According to In all bezel images at each camera angle under each lighting The first The average area of the same layer of the bezel is Under the light The degree of proximity of the area averages of all corresponding bezels of the bezel is obtained. The first Several comparison circles of the same layer circle.
7. The method for processing and inspecting the bezel of a mechanical watch according to claim 1, characterized in that: The specific method of obtaining the final abnormality of the pixel points of each bezel is as follows: Perform edge detection on the bezel to obtain a number of detection areas of each bezel, and obtain a reflective area and a non-reflective area of each bezel according to an average gray value of pixels in the detection area of each bezel; Jordi Under the light The first The corresponding bezel and The first The first If the vibration distances of the first and second comparison bezels are all within the range of their normal vibration distances, the grayscale values of the pixels located in the reflective area, the non-reflective area, and the same position in different areas of the two bezels are fitted respectively, and the grayscale values of the pixels located in the reflective area, the non-reflective area, and the same position in different areas of the two bezels are obtained. A horizontal initial abnormality of the pixel points of the bezel; The product of the defect characteristic variation of the two bezels is taken as the initial lateral abnormality for the first The pixels of the bezel are The influence weight of the lateral anomaly under the illumination is obtained as The first The pixel at The lateral anomaly under each light is calculated to obtain the final anomaly of each pixel point in each bezel.
8. A method for processing and inspecting the bezel of a mechanical watch according to claim 7, characterized in that: The specific method of obtaining the final abnormality of each pixel point of each bezel is as follows: According to the difference between the abscissa of the centroid of each bezel and other bezels in each bezel image, the same column of each bezel is obtained; The first The pixel point at the same position on the circle and all the circles with the same column is The sum of the absolute values of the differences of the lateral anomaly under the first illumination and the The first The pixel at The product of the lateral anomaly under the illumination is taken as the The first The pixel at The degree of abnormal performance under each light; According to the abnormal performance degree of each pixel point of each bezel under each illumination, the final abnormality degree of each pixel point of each bezel is obtained.
9. A method for processing and inspecting the bezel of a mechanical watch according to claim 8, characterized in that: The specific method of obtaining the final abnormality of each pixel point of each bezel is as follows: All pixels of a bezel are The average value of the lateral anomaly under the illumination is taken as the value of the pixel point on the bezel at the The influence weight of the abnormal expression degree under each lighting on the final abnormality degree of the pixel point is calculated to obtain the final abnormality degree of each pixel point in each bezel.
10. A processing quality inspection system for a mechanical watch bezel, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the processing quality inspection method for a mechanical watch bezel as described in any one of claims 1 to 9 are implemented.