A bearing cap defect detection method and detection system

Abstract

This invention belongs to the field of machine vision inspection technology for bearing cover defects, specifically relating to a method and system for detecting bearing cover defects. Addressing the technical problems of misjudgment due to black spots on the bearing cover surface and uneven lighting caused by curved surfaces, this invention employs a dual-station architecture: the first station uses multiple light sources at different angles to individually illuminate the bearing cover, combined with a photometric stereo algorithm to synthesize images, effectively highlighting deep defects at the bottom and suppressing interference from surface black spots, while simultaneously completing the inspection of incoming material orientation and dimensions; the second station, for groups of bearing covers, first uses combined light sources to acquire images for code dot quantity and coding integrity inspection, then switches the light source mode to eliminate curved surface interference and perform placement orientation inspection. This invention links visual inspection with a robotic arm, achieving full automation from material loading and multi-dimensional inspection to the rejection of defective products, significantly improving inspection accuracy and efficiency, and effectively reducing the misjudgment rate and production losses.

Description

Technical Field

[0001] This invention belongs to the field of machine vision inspection technology for bearing cover defects, and specifically relates to a method and system for detecting bearing cover defects. Background Technology

[0002] As a core component in mechanical transmission and bearing assembly, the appearance integrity and standardized marking of bearing caps directly affect subsequent assembly accuracy, product lifecycle traceability management, and overall equipment operational stability. Therefore, in the industrial mass production of bearing caps, conducting efficient and accurate inspection of appearance defects and marking quality is a crucial process for controlling product quality and ensuring the reliability of downstream applications. With the upgrading of manufacturing towards automation and intelligence, machine vision inspection technology, due to its advantages such as high inspection efficiency, non-contact operation, and continuous operation, has gradually replaced manual visual inspection and become the mainstream technology in bearing cap production inspection, widely used in the appearance and marking inspection of various components.

[0003] In the actual production and conveying process of bearing caps, there are situations where bearing caps of various specifications and sizes are transported together, and the placement direction of incoming materials is not standardized. This places basic requirements on specification identification and orientation determination in the inspection process. At the same time, the bottom of the bearing caps is prone to scratches, missing materials, and other appearance defects during processing, transfer, and conveying. These defects directly affect the assembly accuracy and performance of the product, and must be accurately identified and scrapped during the inspection process. In addition, the surface of the bearing caps needs to be marked with a coding machine for traceability management of product production information and batch information. However, coding machines are prone to problems such as missing coding or incomplete coding during continuous operation, and bearing caps placed in groups are also prone to being placed in opposite directions. All these problems require visual inspection technology to make comprehensive judgments from multiple dimensions in the same inspection process, which places high practical application requirements on the comprehensiveness, completeness, and accuracy of the visual inspection solution.

[0004] Existing machine vision inspection solutions for bearing caps, however, fall short of meeting the aforementioned multi-dimensional and precise inspection requirements in practical applications, exhibiting numerous inspection pain points. Due to the material and manufacturing process of the bearing caps themselves, naturally occurring black spots are prone to appear on their surface. Existing visual inspection methods cannot effectively distinguish these natural black spots from actual scratches, missing materials, or other deep defects, easily misjudging black spots as product defects. This results in the incorrect rejection of a large number of qualified products, increasing production losses and reducing overall inspection efficiency due to frequent misjudgments. Furthermore, the upper surface of the bearing cap is not a flat structure but has a certain curvature, causing uneven grayscale in the captured images. This severely interferes with the detection of coding integrity, the number and position of coding dots, leading to coding-related inspection misjudgments and making it difficult to effectively control coding quality.

[0005] The aforementioned problems with existing visual inspection technologies not only significantly reduce the accuracy of detecting appearance defects and marking quality in bearing covers, making it difficult to effectively control product quality before shipment, but also lead to a substantial increase in labor costs for rejecting defective products and re-inspecting qualified products in the production process due to frequent misjudgments. At the same time, the erroneously rejected qualified products also cause unnecessary waste of raw materials. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a method and system for detecting defects in bearing caps. The purpose of this invention is to achieve accurate and efficient detection of appearance defects, marking quality, and placement status of bearing caps, significantly improving the accuracy and reliability of detection judgments, effectively reducing false judgment rates and production losses, and adapting to detection scenarios involving the mixed transportation of multiple product specifications, thereby ensuring the quality of bearing cap products leaving the factory.

[0007] The first aspect of this invention provides a method for detecting defects in a bearing cap, comprising: The first inspection step is to use the first vision inspection unit to acquire multi-angle lighting images of the bottom of the bearing cover, process the acquired images using a photometric stereo algorithm to generate a composite image, and detect deep appearance defects on the bottom of the bearing cover based on the composite image. The second station inspection step: The second vision inspection unit is used to acquire images of the top of multiple bearing covers arranged in groups, and the code detection and placement direction detection are performed based on the acquired images. The first and second inspection steps are executed sequentially. If any inspection step is deemed unqualified, a signal is sent to the actuator to remove the corresponding unqualified product.

[0008] As a further optimization of the above-mentioned bearing cap defect detection method, the first station detection step includes: Image acquisition sub-steps: Control at least four light sources to illuminate the bearing cover individually from different directions and angles, and acquire multiple images of the bottom of the bearing cover with different lighting directions; then control at least four light sources to turn on simultaneously, and acquire a single image with a conventional light source. Bottom appearance defect detection sub-step: Multiple images with different lighting directions are combined into a single composite image using a photometric stereo algorithm. This composite image is then matched with a preset standard template to define the defect detection area. Images within the detection area are filtered and differentially processed. Valid defects are then selected and determined based on preset grayscale and area thresholds.

[0009] As a further optimization of the above-mentioned bearing cover defect detection method, the bottom appearance defect detection sub-step includes filtering and differential processing: performing median filtering on the synthetic image within the defect detection area, performing differential processing on the filtered image and the image before filtering to obtain a differential image; filtering out areas with gray values ​​greater than the gray value threshold in the differential image, and performing morphological closing operation on the filtered areas to identify areas with areas greater than the area threshold as valid defects.

[0010] As a further optimization of the above-mentioned bearing cover defect detection method, the first station detection step also includes a material inlet direction determination sub-step: extracting the notch contour features of the bearing cover image, matching it with a preset direction standard template, calculating the angle deviation between the two, and determining the material inlet direction of the bearing cover based on the angle deviation.

[0011] As a further optimization of the above-mentioned bearing cover defect detection method, the first station detection step also includes a size determination sub-step: using an edge detection algorithm to obtain the upper and lower boundary contour points of the bearing cover from a conventional light source image, calculating the width measurement value of the bearing cover based on the contour points, and comparing the width measurement value with a preset size determination threshold to determine the size of the bearing cover.

[0012] As a further optimization of the above-mentioned bearing cap defect detection method, the second station detection step includes: Positioning sub-step: Based on the acquired images, the center position and angle of each of the multiple bearing covers arranged in groups are located using image algorithms, and the corresponding position coordinate system is established; Code dot quantity arrangement detection sub-step: According to the position coordinate system, delineate multiple code dot detection areas in the image corresponding to the number of bearing covers, find the number of code dots in each area, and determine whether the actual code dot quantity arrangement order is consistent with the preset standard quantity arrangement order. The code integrity detection sub-step is as follows: Based on the position coordinate system, the code integrity detection area is defined at the actual position of each code point, the actual area of ​​the code points within the area is calculated, and the actual area is compared with the preset area threshold to determine whether the code of each code point is complete.

[0013] As a further optimization of the above-mentioned bearing cap defect detection method, the positioning sub-step specifically includes: using a grayscale value lookup method to extract the center position of the circular hole of the bearing cap in the image, connecting the centers of the circular holes of adjacent bearing caps into a straight line, and calculating the angle of the straight line, thereby determining the position and angle of each bearing cap.

[0014] As a further optimization of the above-mentioned bearing cover defect detection method, the second station detection step also includes a product placement orientation detection sub-step: after completing the code point detection, switch to a preset light source mode dedicated to orientation detection for image acquisition; according to the position coordinate system, delineate the orientation detection area in the image, identify the character features on the bearing cover surface within the area, match the character features with a preset orientation standard template, calculate the angle deviation, and if the angle deviation is greater than or equal to a preset angle threshold, it is determined that the bearing cover is placed in the wrong orientation.

[0015] As a further optimization of the above-mentioned bearing cover defect detection method, when performing the product placement direction detection sub-step, the light source mode used is to turn on only the open light source and turn off other light sources used for supplementary lighting, so as to eliminate the light interference caused by the curvature of the bearing cover surface.

[0016] A second aspect of the present invention is to provide a bearing cap defect detection system for implementing the above-mentioned detection method, comprising: A robotic arm is used to grasp and transfer bearing caps; The first visual inspection unit includes a first camera and at least four light sources distributed around the first camera at different azimuth angles, for performing first station inspection; The second vision inspection unit includes a second camera, an open light source, and at least two strip light sources. The strip light sources are used to supplement the light on the side of the bearing cover, and the open light sources are used for backlight illumination during orientation detection. The second vision inspection unit is used to perform the second station inspection. The controller is connected to the robotic arm, the first vision detection unit, and the second vision detection unit respectively. It is used to control image acquisition, image processing, result judgment, and control the robotic arm to remove defective products when they are judged to be unqualified.

[0017] Beneficial effects This invention addresses the technical challenges of misidentifying surface black spots as defects during bearing cover appearance and coding inspection, the uneven lighting caused by the curved surface of the product leading to coding inspection errors, and the need for multi-dimensional inspection of bearing covers of different sizes and with different incoming material orientations. By designing a dual-station differentiated visual inspection architecture, coupled with a customized light source layout, multi-round image acquisition methods, and targeted image algorithm processing strategies, it achieves accurate identification of appearance defects such as scratches and missing materials on the bottom of the bearing cover. It can effectively distinguish between inherent black spots and actual defects with depth. Furthermore, to address the inspection difficulties caused by the curvature of the bearing cover's front, a side-supplementary light source design solves the problem of uneven lighting, achieving high-precision inspection of the number and arrangement of coding dots, coding integrity, and product placement orientation, completely avoiding coding-related inspection errors. In addition... This invention deeply integrates visual inspection processes with automated robotic operations, constructing a fully automated inspection system that spans from product loading and multi-dimensional detection to real-time rejection of defective products and transfer and packaging of qualified products. It effectively rejects bearing caps with poor appearance, inadequate coding, and incorrect placement, and can also work with robotic arms to complete coding operations according to rules, significantly reducing manual intervention and greatly improving the overall efficiency of the bearing cap production inspection process. Furthermore, the relevant parameters of this inspection method can be adaptively adjusted according to the actual bearing cap specifications and precision requirements, exhibiting good inspection adaptability for different specifications of bearing caps. This significantly improves the stability and reliability of inspection judgments, providing efficient and precise technical support for quality control in the bearing cap production process. It optimizes and upgrades bearing cap visual inspection technology in terms of both inspection accuracy and production efficiency. Attached Figure Description

[0018] Figure 1 This is a photograph of the bearing cover taken by a camera under 0° single light source illumination at the first workstation.

[0019] Figure 2 This is a photograph of the bearing cover taken by a camera under 90° single light source illumination at the first workstation.

[0020] Figure 3 This is a photograph of the bearing cover taken by a camera under 180° single-light source illumination at the first workstation.

[0021] Figure 4 This is a photograph of the bearing cover taken by a camera under 270° single light source illumination at the first workstation.

[0022] Figure 5 This is a photograph of the bearing cover taken by a camera under normal lighting conditions at the first workstation.

[0023] Figure 6 This is a composite image of the bearing cover after being synthesized using a photometric stereo algorithm at the first workstation.

[0024] Figure 7This is the image before mid-range filtering in the defect detection area of ​​the composite image of the bearing cover at the first workstation.

[0025] Figure 8 This is the image after medium filtering of the defect detection area in the composite image of the bearing cover at the first workstation.

[0026] Figure 9 This is a differential image of the defect detection area in the composite image of the bearing cover at the first workstation, before and after filtering.

[0027] Figure 10 This is a schematic diagram of the composite image of the bearing cover at the first workstation overlaid with defect markers.

[0028] Figure 11 This is a schematic diagram of the structure of the camera used for inspection at the second workstation, in conjunction with three light sources.

[0029] Figure 12 This is a diagram showing the results of the test for the number of circular code dots arranged in groups of 5 bearing caps at the second workstation.

[0030] Figure 13 This is a display image showing the results of character feature recognition on the surface of the bearing cover at the second workstation.

[0031] Figure 14 This is a schematic diagram of the overall implementation process of the bearing cap defect detection method of the present invention. Detailed Implementation

[0032] This invention discloses a method for detecting defects in bearing caps, applicable to the detection of appearance defects, coding quality, and placement status of bearing caps of two sizes and specifications transported by belt conveyors. It solves the technical problems of misjudging black spots on the bearing cap surface as defects and uneven grayscale in curved surface coding detection caused by traditional visual inspection. The method employs a dual-camera, dual-station visual inspection structure in conjunction with a robotic arm to complete the entire inspection process. The first station detects the bearing cap's incoming material direction, size, and bottom appearance defects. The second station detects the coding integrity, code dot quantity arrangement, and product placement direction of five bearing caps per group. The specific implementation of this invention is described in detail below.

[0033] The first station is equipped with an inspection camera and four light sources to determine the incoming direction and size of the bearing cover, as well as to detect appearance defects such as bottom scratches and missing materials. The main steps include image acquisition, bottom appearance defect detection, incoming direction determination, and size determination.

[0034] The image acquisition process involves distributing four light sources at azimuth angles of 0°, 90°, 180°, and 270°, with all light sources set to an illumination angle of 45° and their brightness adjusted to 255. Each of the four light sources is then individually controlled to illuminate the image, and the camera takes four corresponding images. Figures 1 to 4As shown, four images of the bottom of the bearing cover were obtained; after four single-light source shots, all four light sources were turned on simultaneously to form a conventional light source, and the camera took another shot to obtain an image as shown. Figure 5 The image shown is a standard image of the bearing cap, used for subsequent dimensional inspection.

[0035] Bottom appearance defect detection uses a photometric stereo algorithm to combine the four images taken by a single light source into a single image, such as... Figure 6 As shown, deep defects such as scratches and missing materials at the bottom of the bearing cover are clearly displayed. A standard template of the bearing cover is pre-made, and a defect detection area is marked on the template. A template matching algorithm is used to calculate the template position and angle between the subsequent image to be detected and the standard template, and to determine the corresponding defect detection area position in the image to be detected. The composite image within the defect detection area is subjected to median filtering. The images before and after filtering are shown below. Figure 7 and Figure 8 As shown, the filtered image and the unfiltered image are subjected to a difference process, and the resulting image is shown below. Figure 9 As shown; a grayscale threshold is preset, and regions with grayscale values ​​greater than the threshold are selected within the defined detection area; morphological closing operations are performed on the selected regions to eliminate holes within the regions and connect adjacent pixels; then a region area threshold is preset, and only regions with an area greater than the area threshold are considered valid defects on the bottom of the bearing cover, while regions with grayscale values ​​exceeding the threshold but areas below the threshold are considered non-defects. Figure 10 The image shown is a schematic diagram of the overlay of defect markers on a composite image.

[0036] The incoming material direction determination adopts a template matching algorithm. A bearing cover image is selected, its notch contour is extracted and made into a standard template for direction determination; feature recognition is performed on the image to be detected, and the angle deviation between the recognized notch feature and the notch feature in the standard template is calculated. The incoming material direction of the bearing cover is determined based on the angle deviation.

[0037] When determining the size, for images captured by conventional light sources, a caliper tool combined with an edge detection algorithm is used to identify and obtain the coordinates of the upper and lower boundary contour points of the bearing cover image. The least squares method is used to fit all contour points of the lower boundary into a straight line, and the distance from each contour point of the upper boundary to the fitted straight line is calculated. The average of all distance values ​​is taken as the width of the bearing cover measured in this test. It is known that the pixel deviation of the upper and lower boundary widths of large and small bearing covers in the camera's field of view is 30 pixels. The middle pixel value of the width of the large and small products is taken as the size determination threshold. If the measured width of the product to be tested is less than the threshold, it is determined to be a small bearing cover; if it is greater than the threshold, it is determined to be a large bearing cover.

[0038] The second workstation is equipped with one inspection camera and three light sources, such as Figure 11As shown, it includes an open light source and two strip light sources. The two strip light sources are used to supplement the side of the bearing cover, adapting to the curvature features of the front of the bearing cover, and completing the detection of the number of circular code dots in a group of 5 bearing covers, the integrity of the coding, and the product placement direction. The main steps include code dot number arrangement detection, coding integrity detection, and product placement direction detection.

[0039] During the code dot quantity arrangement detection, the open light source and two bar light sources are turned on simultaneously. A grayscale value lookup method is used to extract the center positions of the circular holes on both sides of the five bearing cover images to be detected. The found circular holes are then connected in pairs from top to bottom according to coordinate order, and the angle of each pair of lines is calculated. This method determines the position and angle coordinate system corresponding to the five bearing cover images. Based on these five coordinate systems, five code dot quantity detection areas are defined in the image to be detected. Within each detection area, a template matching method is used to find the number of circular code dots. It is then determined whether the number of code dots detected from top to bottom in the five detection areas is sequentially 4, 3, 2, 1, 0. The results are as follows: Figure 12 As shown; if the arrangement order does not match the quantity sequence, it is determined that the code points are missing or over-printed, and the code quantity arrangement is not qualified; if it matches the quantity sequence, it is determined that the code point quantity arrangement is qualified.

[0040] During the completeness inspection of the coding, based on the above 5 sets of coordinate systems, one coding integrity inspection area is defined for each actual position of the coding dot, for a total of 10 inspection areas. The actual area of ​​the circular coding dot in each inspection area is calculated using a grayscale thresholding method. Morphological closing operations are performed on the coding dot image in each inspection area to eliminate holes in the coding dot area. A coding dot area threshold is preset. If the actual area of ​​the coding dot in the inspection area is greater than the area threshold, the coding dot is judged to be complete. If the actual area is less than the area threshold, the coding dot is judged to be defective, and the coding integrity is unqualified.

[0041] When detecting the product placement orientation, two strip light sources are turned off, leaving only the backlight on to eliminate light interference from the curvature of the bearing cover's front. A template matching algorithm is used to select images of normally placed bearing covers, extract their surface character outlines, and create a standard template for orientation determination. Based on the previously determined five coordinate systems, five orientation detection areas are defined in groups of five bearing cover images to be detected. Character features within each detection area are then identified, such as... Figure 13 As shown, calculate the angular deviation between the bearing cover and the character features in the standard template. If the angular deviation is less than 90 degrees, the bearing cover is considered to be placed correctly. If the angular deviation is greater than or equal to 90 degrees, the bearing cover is considered to be placed backwards.

[0042] The bearing cap defect detection method of this invention, in conjunction with a robotic arm, completes the entire process of product loading, inspection and judgment, rejection of defective products, and transfer of qualified products. The overall implementation process is as follows: Figure 14 As shown, it includes the following steps: Step S1: The robotic arm picks up the bearing cover to be inspected and places it above the camera at the first station. Step S2: The camera at the first station completes four single-light source shots and one conventional light source shot according to the aforementioned steps, sequentially inspecting the appearance defects, incoming material direction, and size of the bearing cover bottom; if any inspection item is deemed unqualified, an NG signal is immediately sent to the robot arm, which then removes the unqualified bearing cover; if all inspection items are deemed qualified, the process proceeds to the next inspection step. Step S3: The robotic arm transfers the bearing covers that have passed the inspection at the first station to the shooting area of ​​the camera at the second station, arranging them in groups of 5; Step S4: The second station camera first turns on the aperture light source and two bar light sources to sequentially complete the code dot quantity arrangement and coding integrity detection; if any detection item is determined to be unqualified, an NG signal is immediately sent to the robot, and the robot removes the group of unqualified products; if all coding-related detection items are determined to be qualified, proceed to the next detection process. Step S5: The second station camera turns off the two strip light sources and turns on only the backlight to complete the placement orientation detection of 5 bearing covers per group; if any product is found to be placed backwards, an NG signal is immediately sent to the robot, which then removes the unqualified products from the group; if the placement orientation of all products is determined to be normal, then all inspection items of the bearing covers in that group are qualified. Step S6: The robotic arm transfers all the qualified bearing caps from the second workstation to the designated area for subsequent packaging.

[0043] The tools or algorithms not described in detail in this embodiment are all conventional technical means in the field of machine vision inspection. Their specific parameter settings can be adaptively adjusted according to the actual bearing cover specifications, camera resolution, and inspection accuracy requirements.

[0044] The above embodiments are exemplary and are intended to illustrate the technical concept and features of the present invention, so that those skilled in the art can understand the content of the present invention and implement it accordingly. They should not be construed as limiting the scope of protection of the present invention. All equivalent changes or modifications made according to the spirit and essence of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for detecting defects in a bearing cap, characterized in that, include: The first station inspection step: The first vision inspection unit is used to collect multi-angle lighting images of the bottom of the bearing cover. The collected images are processed by a photometric stereo algorithm to generate a composite image. The composite image is then used to detect deep appearance defects on the bottom of the bearing cover. The second station inspection step: The second vision inspection unit is used to acquire images of the top of multiple bearing covers arranged in groups, and the code detection and placement direction detection are performed based on the acquired images. The first station inspection step and the second station inspection step are executed sequentially. If any inspection step is determined to be unqualified, a signal is sent to the execution mechanism to remove the corresponding unqualified product.

2. The bearing cap defect detection method according to claim 1, characterized in that, The first station inspection steps include: Image acquisition sub-step: Control at least four light sources to illuminate the bearing cover individually from different directions and angles, and acquire multiple images of the bottom of the bearing cover with different lighting directions; then control the at least four light sources to turn on simultaneously, and acquire a conventional light source image; Bottom appearance defect detection sub-step: The multiple images with different lighting directions are combined into a single composite image using a photometric stereo algorithm. The composite image is then matched with a preset standard template to define the defect detection area. The images within the detection area are filtered and differentially processed. Valid defects are then screened and determined based on preset grayscale and area thresholds.

3. The bearing cap defect detection method according to claim 2, characterized in that, In the bottom appearance defect detection sub-step, the filtering and differential processing includes: performing median filtering on the synthetic image within the defect detection area, performing differential processing on the filtered image and the image before filtering to obtain a differential image; filtering out areas with gray values ​​greater than the gray value threshold in the differential image, and performing morphological closing operation on the filtered areas to identify areas with areas larger than the area threshold as valid defects.

4. The bearing cap defect detection method according to claim 2, characterized in that, The first station inspection step also includes a material inlet direction determination sub-step: extracting the notch contour features of the bearing cover image, matching it with a preset direction standard template, calculating the angle deviation between the two, and determining the material inlet direction of the bearing cover based on the angle deviation.

5. The bearing cap defect detection method according to claim 2, characterized in that, The first station detection step also includes a size determination sub-step: using an edge detection algorithm to obtain the upper and lower boundary contour points of the bearing cover from the conventional light source image, calculating the width measurement value of the bearing cover based on the contour points, and comparing the width measurement value with a preset size determination threshold to determine the size of the bearing cover.

6. The bearing cap defect detection method according to claim 1, characterized in that, The second station inspection step includes: Positioning sub-step: Based on the acquired images, the center position and angle of each of the multiple bearing covers arranged in groups are located using image algorithms, and the corresponding position coordinate system is established; Code dot quantity arrangement detection sub-step: According to the position coordinate system, delineate multiple code dot detection areas in the image corresponding to the number of bearing covers, find the number of code dots in each area, and determine whether the actual code dot quantity arrangement order is consistent with the preset standard quantity arrangement order. The code integrity detection sub-step is as follows: Based on the position coordinate system, a code integrity detection area is defined at the actual position of each code point, the actual area of ​​the code points within the area is calculated, and the actual area is compared with a preset area threshold to determine whether the code of each code point is complete.

7. The bearing cap defect detection method according to claim 6, characterized in that, The positioning sub-step specifically includes: using a grayscale value lookup method to extract the center position of the circular hole of the bearing cover in the image, connecting the centers of the circular holes of adjacent bearing covers into a straight line, and calculating the angle of the straight line, thereby determining the position and angle of each bearing cover.

8. The bearing cap defect detection method according to claim 6, characterized in that, The second station detection step also includes a product placement orientation detection sub-step: after completing the code point detection, switch to a preset light source mode dedicated to orientation detection for image acquisition; according to the position coordinate system, delineate the orientation detection area in the image, identify the character features on the bearing cover surface within the area, match the character features with a preset orientation standard template, calculate the angle deviation, and if the angle deviation is greater than or equal to a preset angle threshold, determine that the bearing cover is placed in the wrong orientation.

9. The bearing cap defect detection method according to claim 8, characterized in that, When performing the product placement orientation detection sub-step, the light source mode used is to turn on only the open light source and turn off other light sources used for supplementary lighting in order to eliminate light interference caused by the curvature of the bearing cover surface.

10. A bearing cap defect detection system, used to implement the method as described in any one of claims 1 to 9, characterized in that, include: A robotic arm is used to grasp and transfer bearing caps; The first visual inspection unit includes a first camera and at least four light sources distributed around the first camera at different azimuth angles, for performing first station inspection; The second vision inspection unit includes a second camera, an open light source, and at least two strip light sources. The strip light sources are used to provide supplementary lighting for the side of the bearing cover, and the open light sources are used for backlight illumination during orientation detection. The second vision inspection unit is used to perform second-station inspection. The controller is connected to the robotic arm, the first vision detection unit, and the second vision detection unit respectively, and is used to control image acquisition, image processing, result judgment, and control the robotic arm to remove defective products when they are judged to be unqualified.

Images