A wheel hub surface detection method based on machine vision

Abstract

This invention relates to the field of quality inspection technology, and more particularly to a machine vision-based method for inspecting the surface of wheel hubs. The method includes: setting three image acquisition devices at specific locations on the end face of the wheel hub to acquire images and preprocess them to extract a first image of defects; if a defect image is acquired, then second and third images of the defects are obtained. Based on multiple defect images, changing feature values ​​are calculated to initially distinguish between raised and non-raised defects. For non-raised defects, the specific category is determined based on the overlap of feature marker points. This invention uses multi-view image acquisition for comprehensive judgment, employs a hierarchical judgment method for accurate classification, targeted preprocessing for rapid defect separation, and reasonable selection of feature values ​​to provide reliable evidence. It optimizes the inspection process, not only identifying defects but also clarifying their types, providing a basis for subsequent processing and improving the level of inspection technology.

Description

Technical Field

[0001] This invention relates to the field of quality inspection technology, and in particular to a method for inspecting wheel hub surfaces based on machine vision. Background Technology

[0002] As a critical safety component, the surface defects of automotive wheel hubs directly affect the overall aesthetics and driving safety of the vehicle. Traditional manual inspection methods suffer from low efficiency, high missed detection rates, and strong subjectivity, and can no longer meet the demands of modern production lines for high-speed, high-precision, and all-weather inspection.

[0003] With the advancement of technology, machine vision-based intelligent inspection technology is gradually replacing manual inspection and becoming the mainstream approach. This technology integrates high-resolution imaging systems, light source control, image processing algorithms, and deep learning models to achieve automatic identification, location, and classification of defects on wheel hub surfaces. However, a problem arises where subsequent correction of defects is impossible due to the inability to distinguish between different defect categories. Therefore, there is an urgent need for a machine vision-based wheel hub surface inspection method to classify and process defects in automotive wheel hubs. Summary of the Invention

[0004] Therefore, the present invention provides a machine vision-based wheel hub surface inspection method to overcome the problem that existing technologies cannot classify and process defects in automobile wheel hubs.

[0005] To achieve the above objectives, the present invention provides a system comprising:

[0006] Step S1: Along the end face of the wheel hub, with the midpoint of the radius of any wheel hub end face as the target acquisition point, a first image acquisition device, a second image acquisition device, and a third image acquisition device are respectively set up based on the vertical plane of the target acquisition point, wherein the first image acquisition device is perpendicular to the end face of the wheel hub.

[0007] Step S2: Turn on the first image acquisition device to acquire the target detection area of ​​the wheel hub to obtain a first image, and preprocess the first image to extract the defect image to obtain the first defect image;

[0008] Step S3: In response to the first image acquisition device acquiring a defect image, the second image acquisition device and the third image acquisition device are turned on respectively to obtain a second defect image and a third defect image;

[0009] Step S4: Based on the first defect image, the second defect image, and the third defect image, obtain the defect image change feature value and preliminarily determine the defect category according to the defect image change feature value. The defect category includes convex defects and non-convex defects.

[0010] Step S5: In response to the preliminary determination that the defect is a non-protruding defect, the feature marker points of the first image of the defect, the second image of the defect, and the third image of the defect are obtained respectively, and the category of the non-protruding defect is determined according to the overlap of the feature marker points. The categories of non-protruding defects include planar defects and pit defects.

[0011] Step S6: Mark each detected defect, and rotate the wheel hub to perform defect detection in the next target detection area.

[0012] Further, in step S2, the process of acquiring the first image of the defect includes:

[0013] Obtain the brightness of each pixel in the first image;

[0014] Mark pixels with brightness less than or equal to the preset brightness;

[0015] The image composed of the marked pixels is output as the first defect image.

[0016] Furthermore, the process of acquiring the second image or the third image of the defect includes:

[0017] Obtain the position and brightness of the reference pixel in the first image;

[0018] Obtain the position and brightness of a reference pixel in the second and / or third image;

[0019] Align the brightness of each pixel in the second and / or third images based on the position and brightness of the reference pixel in the first image;

[0020] Mark the pixels in the second and / or third images whose brightness is less than or equal to the preset brightness;

[0021] Output the image composed of the marked pixels as the second defect image and / or the third defect image.

[0022] Furthermore, in step S3, in response to the first image acquisition device failing to acquire a defect image, the wheel hub is rotated to perform defect detection in the next target detection area.

[0023] Furthermore, the defect image change feature value is the ratio of the area of ​​the union image formed by the first defect image, the second defect image, and the third defect image to the area of ​​the first defect image.

[0024] Further, in step S4, in response to the defect image change feature value being greater than a preset defect image change characterization value, the defect category is determined to be the protrusion type defect.

[0025] Further, in step S4, in response to the defect image change feature value being equal to the preset defect image change characterization value, the defect category is determined to be the non-protrusion type defect.

[0026] Furthermore, the pixels with the lowest brightness in the first defect image, the second defect image, and the third defect image are respectively marked as the first feature marker point, the second feature marker point, and the third feature marker point.

[0027] Furthermore, the overlap of the feature marker points is the average distance between the first feature marker point, the second feature marker point, and the third feature marker point.

[0028] Furthermore, the category of non-protruding defects is determined based on the degree of overlap of each of the aforementioned feature marker points, wherein,

[0029] If the overlap of the feature marker points is less than the preset overlap threshold of the feature marker points, then the defect category is determined to be a depression defect.

[0030] If the overlap of the feature marker points is equal to the preset overlap threshold of the feature marker points, then the defect category is determined to be a planar defect.

[0031] Compared with the prior art, the beneficial effects of the present invention are as follows: by setting up three image acquisition devices to acquire images of the wheel hub surface from different angles, and using multi-view image information to comprehensively judge defects, the defect features can be obtained more comprehensively and accurately than single-view detection, thereby improving the accuracy and reliability of defect detection.

[0032] Furthermore, a hierarchical method for classifying defects is adopted. First, based on the feature values ​​of defect image changes, convex and non-convex defects are initially distinguished. Then, for non-convex defects, the specific category is determined based on the overlap of feature marker points. This hierarchical method has a clear logic and can classify various types of defects more accurately.

[0033] Furthermore, during the preprocessing of the first image, pixels different from those in the reference image of the wheel hub surface are extracted to determine the defect contour. This targeted preprocessing method helps to quickly and accurately separate the defect parts from the image, thereby improving the defect detection efficiency.

[0034] Furthermore, the ratio of the union of the shadow areas of the defect contours in the first, second, and third defect images to the shadow area of ​​the defect contour in the first defect image is used as the defect image change feature value. The pixels with the darkest brightness in the image are used as feature markers. The selection of these feature values ​​is reasonable and representative, and can effectively reflect the characteristics of the defect, providing a reliable basis for accurately determining the defect category.

[0035] Furthermore, when the first image acquisition device fails to acquire a defect image, the steps of marking and rotating the hub to proceed to the next area for detection are directly executed, avoiding unnecessary operations, optimizing the detection process, and improving the overall detection efficiency.

[0036] This invention utilizes multiple image acquisition devices working collaboratively, combined with image processing and feature extraction techniques. This not only enables rapid identification of defects but also clarifies their specific types, providing a scientific basis for subsequent repair or treatment. This proposed method elevates the technical level of wheel hub surface defect detection, solving the problems of low efficiency and high false positive rates inherent in traditional detection methods. Attached Figure Description

[0037] Figure 1 This is a flowchart of a machine vision-based wheel hub surface detection method according to an embodiment of the present invention;

[0038] Figure 2 This is a flowchart illustrating the process of initially determining the defect category based on the feature values ​​of defect image changes, as described in an embodiment of the present invention.

[0039] Figure 3 This is a flowchart illustrating how the category of non-protruding defects is determined based on the overlap of each feature marker point in an embodiment of the present invention.

[0040] Figure 4 This is a schematic diagram of a machine vision-based wheel hub surface detection method according to an embodiment of the present invention;

[0041] In the diagram: 1. First image acquisition device; 2. Second image acquisition device; 3. Third image acquisition device. Detailed Implementation

[0042] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0043] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0044] Please see Figure 1-4 The figures shown are, respectively, a flowchart of a wheel hub surface detection method based on machine vision according to an embodiment of the present invention; a flowchart of a preliminary determination of the defect category based on the feature value of defect image changes according to an embodiment of the present invention; a flowchart of a determination of the category of non-protruding defects based on the overlap of each feature marker point according to an embodiment of the present invention; and a structural schematic diagram of a wheel hub surface detection method based on machine vision according to an embodiment of the present invention.

[0045] An embodiment of the present invention provides a machine vision-based wheel hub surface detection method, comprising:

[0046] Step S1: Using the midpoint of any radius of the wheel hub end face as the target acquisition point, set up a first image acquisition device 1, a second image acquisition device 2, and a third image acquisition device 3 on the vertical plane of the target acquisition point. The first image acquisition device 1 is perpendicular to the wheel hub end face; the second image acquisition device 2 and the third image acquisition device 3 are symmetrically set at both ends of the first image acquisition device 1 with an included angle of 30°.

[0047] Step S2: Turn on the first image acquisition device 1 to acquire the target detection area of ​​the wheel hub to obtain a first image, and preprocess the first image to extract the defect image to obtain the first defect image; during the preprocessing process, the first image is grayscaled to reduce the amount of data and retain the key information of the defect.

[0048] Step S3: In response to the first image acquisition device 1 acquiring a defect image, the second image acquisition device 2 and the third image acquisition device 3 are activated respectively to obtain a second defect image and a third defect image;

[0049] Step S4: Based on the first defect image, the second defect image, and the third defect image, obtain the defect image change feature value and preliminarily determine the defect category according to the defect image change feature value. The defect category includes convex defects and non-convex defects.

[0050] Step S5: In response to the preliminary determination that the defect is a non-protruding defect, the feature marker points of the first image of the defect, the second image of the defect, and the third image of the defect are obtained respectively, and the category of the non-protruding defect is determined according to the overlap of the feature marker points. The categories of non-protruding defects include planar defects and pit defects.

[0051] Step S6: Mark each detected defect, and rotate the wheel hub to perform defect detection in the next target detection area.

[0052] Specifically, in step S1, the position of the first image acquisition device 1 is set as the core reference point, which is perpendicular to the end face of the wheel hub to ensure that the acquired image has the highest clarity and accuracy. The second image acquisition device 2 and the third image acquisition device 3 are symmetrically distributed on both sides of the first image acquisition device 1 at a 30° angle. This layout can effectively cover different angles of the wheel hub surface and reduce blind spots caused by a single viewpoint. The first image acquisition device 1, the second image acquisition device 2, and the third image acquisition device 3 can be industrial cameras, and there is no specific limitation, as long as they can acquire images of the target detection area of ​​the wheel hub.

[0053] Specifically, by adjusting the relative positions and angles between the three image acquisition units, the comprehensiveness and accuracy of image acquisition can be further optimized, providing more reliable data support for subsequent defect analysis.

[0054] Specifically, in actual operation, a high-precision mounting bracket is used for the installation and positioning of the image acquisition device to ensure its stability. Simultaneously, to avoid interference from ambient light on image quality, a dedicated light source system is provided in the detection area. By adjusting the intensity and angle of the light source, the acquired image is ensured to have uniform brightness and no obvious shadows. This collaborative design between the light source and the image acquisition device not only improves the quality of image acquisition but also significantly reduces the misjudgment rate caused by external factors.

[0055] Specifically, before the image acquisition devices operate, they need to be calibrated to ensure that the imaging parameters of each acquisition device are consistent. The calibration process includes uniformly setting the focal length, aperture size, and exposure time, thereby ensuring that the three sets of images have a high degree of consistency in terms of brightness, contrast, and resolution. This rigorous process lays a solid foundation for subsequent image processing and feature extraction, further improving the overall performance of the detection system.

[0056] Specifically, in step S2, the process of obtaining the first image of the defect includes:

[0057] Obtain the brightness of each pixel in the first image;

[0058] Mark pixels with brightness less than or equal to the preset brightness;

[0059] The image composed of the marked pixels is output as the first defect image. The preset brightness target value is set to 60% grayscale. It should be noted that the data in this embodiment are all results obtained through preliminary experiments before the system was run. Each preset value can be adjusted according to specific usage, as long as the system method can clearly define different specific situations in the single-item judgment process through the acquired values. Subsequently, a Gaussian filtering algorithm is used to remove interference from the image, ensuring the accuracy of subsequent feature extraction. This process not only effectively removes background interference but also enhances the identifiability of the defect area, laying the foundation for subsequent multi-view image analysis.

[0060] Specifically, the process of obtaining the second image or the third image of the defect includes:

[0061] Obtain the position and brightness of the reference pixel in the first image;

[0062] Obtain the position and brightness of a reference pixel in the second and / or third image;

[0063] Align the brightness of each pixel in the second and / or third images based on the position and brightness of the reference pixel in the first image;

[0064] Pixels in the second and / or third images whose brightness is less than or equal to a preset brightness are marked; wherein, the position of the reference pixel in the first, second and / or third images is the midpoint of the hub radius, and the target brightness value is 60% grayscale;

[0065] Output the image composed of the marked pixels as the second defect image and / or the third defect image.

[0066] Specifically, by acquiring and utilizing the positional information of reference pixels for image alignment, positional deviations caused by factors such as shooting angle and minor object movements between different images can be effectively eliminated. This allows for a more accurate comparison of pixel brightness at corresponding positions in each image during subsequent defect detection, avoiding misjudgments due to positional mismatches and thus improving the accuracy of defect detection.

[0067] Specifically, in step S3, in response to the first image acquisition device 1 failing to acquire a defect image, the wheel hub is rotated to perform defect detection in the next target detection area.

[0068] Specifically, in step S3, when the first image acquisition device 1 fails to acquire a defect image, the defect detection of the next target detection area is performed by rotating the wheel hub. This design ensures that all areas of the wheel hub surface can be effectively detected, avoiding blind spots caused by limitations of a single acquisition angle or position, greatly improving the comprehensiveness and completeness of wheel hub surface defect detection, and guaranteeing the reliability of the overall quality inspection of the wheel hub.

[0069] Specifically, the defect image change feature value is the ratio of the area of ​​the union image formed by the first defect image, the second defect image, and the third defect image to the area of ​​the first defect image.

[0070] Specifically, in step S4, the defect category is determined to be the protrusion type defect in response to the defect image change feature value being greater than the preset defect image change characterization value; wherein, the preset defect image change characterization value is set to 1.06.

[0071] Specifically, in step S4, in response to the defect image change feature value being equal to the preset defect image change characterization value, the defect category is determined to be the non-protrusion type defect.

[0072] Specifically, the pixels with the lowest brightness in the first defect image, the second defect image, and the third defect image are respectively marked as the first feature marker point, the second feature marker point, and the third feature marker point.

[0073] Specifically, marking the pixels with the lowest brightness in a defect image as feature markers can accurately pinpoint the locations of the most significant brightness changes within the defect region. These locations are often the key features of the defect. By analyzing the overlap of these feature markers, we can gain a deeper understanding of how defects behave and change across different images, providing strong feature support for defect classification.

[0074] Specifically, the overlap of the feature marker points is the average distance between the first feature marker point, the second feature marker point, and the third feature marker point, wherein the first defect image, the second defect image, and the third defect image are made to overlap by the position of the reference pixel point.

[0075] Specifically, the category of non-protruding defects is determined based on the degree of overlap of each of the aforementioned feature marker points, wherein,

[0076] If the overlap of the feature marker points is less than the preset overlap threshold of the feature marker points, then the defect category is determined to be a depression defect.

[0077] If the overlap of the feature marker points is equal to the preset overlap threshold of the feature marker points, then the defect category is determined to be a planar defect; wherein, the preset overlap threshold of the feature marker points is set to 0.54 mm.

[0078] Specifically, the overlap of feature marker points reflects the relative spatial positions of key defect features in different images. This setting allows for the acquisition of spatial distribution information of defects on the wheel hub surface, thus enabling a more accurate determination of the defect type and nature. For example, a low overlap of feature marker points for a dent defect indicates significant positional variation across different images, consistent with the spatial characteristics of a dent defect; while a feature marker point overlap of a planar defect equal to a preset threshold indicates relatively stable position across different images, consistent with the characteristics of a planar defect.

[0079] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0080] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A machine vision-based method for inspecting the surface of a wheel hub, characterized in that, include: Step S1: Along the end face of the wheel hub, with the midpoint of the radius of any wheel hub end face as the target acquisition point, a first image acquisition device, a second image acquisition device, and a third image acquisition device are respectively set up based on the vertical plane of the target acquisition point, wherein the first image acquisition device is perpendicular to the end face of the wheel hub. Step S2: Turn on the first image acquisition device to acquire the target detection area of ​​the wheel hub to obtain a first image, and preprocess the first image to extract the defect image to obtain the first defect image; Step S3: In response to the first image acquisition device acquiring a defect image, the second image acquisition device and the third image acquisition device are turned on respectively to obtain a second defect image and a third defect image; Step S4: Based on the first defect image, the second defect image, and the third defect image, obtain the defect image change feature value and preliminarily determine the defect category according to the defect image change feature value. The defect category includes convex defects and non-convex defects. Step S5: In response to the preliminary determination that the defect is a non-protruding defect, the feature marker points of the first image of the defect, the second image of the defect, and the third image of the defect are obtained respectively, and the category of the non-protruding defect is determined according to the overlap of each feature marker point. The categories of non-protruding defects include planar defects and pit defects. Step S6: Mark each detected defect, and rotate the wheel hub to perform defect detection in the next target detection area; Specifically, the pixels with the lowest brightness in the first defect image, the second defect image, and the third defect image are respectively marked as the first feature marker point, the second feature marker point, and the third feature marker point.

2. The machine vision-based wheel hub surface inspection method according to claim 1, characterized in that, In step S2, the process of obtaining the first image of the defect includes: Obtain the brightness of each pixel in the first image; Mark pixels with brightness less than or equal to the preset brightness; The image composed of the marked pixels is output as the first defect image.

3. The machine vision-based wheel hub surface inspection method according to claim 2, characterized in that, The process of obtaining the second image or the third image of the defect includes: Obtain the position and brightness of the reference pixel in the first image; Obtain the position and brightness of a reference pixel in the second and / or third image; Align the brightness of each pixel in the second and / or third images based on the position and brightness of the reference pixel in the first image; Mark the pixels in the second and / or third images whose brightness is less than or equal to a preset brightness; Output the image composed of the marked pixels as the second defect image and / or the third defect image.

4. The machine vision-based wheel hub surface inspection method according to claim 3, characterized in that, In step S3, in response to the first image acquisition device failing to acquire a defect image, the wheel hub is rotated to perform defect detection in the next target detection area.

5. The machine vision-based wheel hub surface inspection method according to claim 4, characterized in that, The defect image change feature value is the ratio of the area of ​​the union image formed by the first defect image, the second defect image, and the third defect image to the area of ​​the first defect image.

6. The machine vision-based wheel hub surface inspection method according to claim 5, characterized in that, In step S4, in response to the defect image change feature value being greater than the preset defect image change characterization value, the defect category is determined to be the protrusion type defect.

7. The machine vision-based wheel hub surface inspection method according to claim 6, characterized in that, In step S4, in response to the defect image change feature value being equal to the preset defect image change characterization value, the defect category is determined to be the non-protrusion type defect.

8. The machine vision-based wheel hub surface inspection method according to claim 7, characterized in that, The overlap of the feature marker points is the average distance between the first feature marker point, the second feature marker point, and the third feature marker point.

9. The machine vision-based wheel hub surface inspection method according to claim 8, characterized in that, The category of non-protruding defects is determined based on the degree of overlap of the aforementioned feature marker points, wherein, If the overlap of the feature marker points is less than the preset overlap threshold of the feature marker points, then the defect category is determined to be a depression defect. If the overlap of the feature marker points is equal to the preset overlap threshold of the feature marker points, then the defect category is determined to be a planar defect.

Images