Semitrailer axle defect identification method based on computer vision
By constructing a multi-dimensional evaluation method for crack coefficient, depression coefficient and roughness coefficient, the problems of single feature reliance and insufficient detection accuracy in existing technologies are solved, and accurate quantitative evaluation and automated closed-loop management of semi-trailer axle defects are achieved, thereby improving detection accuracy and production efficiency.
Patent Information
- Application Number
- CN202510839537.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-23
- Publication Date
- 2025-09-19
AI Technical Summary
Existing computer vision inspection technology has problems in identifying defects in semi-trailer axles, such as single feature reliance, insufficient detection accuracy, and lack of closed-loop management, making it difficult to achieve multi-dimensional feature fusion and automated closed-loop management.
Using a computer vision-based method, through image data acquisition, feature extraction, feature analysis and comprehensive processing, the crack coefficient, sag coefficient and roughness coefficient are constructed to form a multi-dimensional defect assessment, and automated detection and closed-loop management are carried out in combination with the production process.
It has achieved accurate quantitative evaluation and multi-dimensional characterization of semi-trailer axle defects, improved detection accuracy and automated coordination of production processes, and reduced defect incidence and repair costs.
Smart Images
Figure CN120672737A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of semi-trailer axle quality detection, and in particular to a semi-trailer axle defect recognition method based on computer vision. Background Art
[0002] The semi-trailer axle is a core component of the vehicle's load-bearing system. Surface defects on the axle tube (such as cracks, protrusions, depressions, holes, burrs, etc.) directly affect the vehicle's load transfer efficiency and driving safety.
[0003] Although existing computer vision detection technology can partially replace manual labor, its limitations are: Single feature reliance: Most methods target only a single defect type (e.g., detecting only cracks or analyzing only contours), and fail to form a comprehensive evaluation model that integrates multi-dimensional features. Insufficient detection accuracy: The ability to identify defects in complex lighting environments (such as low-contrast scratches and hidden dents) is limited, and the pre-processing process lacks an adaptive adjustment mechanism;
[0004] Lack of closed-loop management: The test results are not deeply linked with the production process, and the automated closed loop of "testing, analysis, and process optimization" cannot be achieved, making it difficult to reduce the defect rate from the root.
[0005] Therefore, there is an urgent need for a computer vision-based method for identifying defects in semi-trailer axles to achieve quantitative evaluation, and an automated detection method that is deeply coordinated with the production process to improve the quality control level and production efficiency of semi-trailer axles. Summary of the Invention
[0006] The purpose of the present invention is to solve the above problems and to propose a method for identifying axle defects of a semi-trailer based on computer vision.
[0007] In order to achieve the above object, the present invention adopts the following technical solutions: A method for identifying axle defects of a semi-trailer based on computer vision, comprising: Image data acquisition: obtain the axle tube image information of the axle; Feature extraction: After pre-processing the image information of the shaft tube, the crack and scratch features, convex and concave features, and hole and burr features of the shaft tube are extracted; Feature analysis: After analyzing the crack and scratch features, protrusion and depression features, and hole burr features, the crack coefficient, protrusion coefficient, and roughness coefficient are obtained; Comprehensive processing: The crack coefficient, depression coefficient and roughness coefficient are processed comprehensively to obtain the defect assessment coefficient; Defect level assessment: Match the corresponding defect level based on the defect assessment coefficient and perform corresponding processing.
[0008] Preferably, the image data acquisition specifically includes: The shaft tube is photographed in real time from multiple angles using fixed-position cameras and light sources, including: The camera is an industrial camera with a frame rate that matches the production line speed and supports line scan; The light source includes: backlight and multi-angle ring light source.
[0009] Preferably, the process of preprocessing the image information of the shaft tube specifically includes: Convert the color image into a grayscale image to eliminate color interference; then use Gaussian filtering or median filtering to remove random noise and salt and pepper noise and smooth the image; Use histogram equalization or adaptive histogram equalization to improve image contrast and enhance the difference between defects and background; sharpen images through Laplacian operator or high-pass filtering to highlight edge details; Finally, illumination compensation is performed to eliminate the effects of uneven illumination, ensuring that cracks, scratches, bumps, depressions, holes, and burr defects on the central axis tube of the image are clearly discernible while maintaining the consistency of the overall image quality.
[0010] Preferably, the process of obtaining the crack coefficient includes: According to the preset size and area, the axle tube is evenly divided into regions to obtain various judgment sub-regions; Obtain the crack features and scratch features of each judgment sub-region, and connect the beginning and end points of each crack and scratch with a straight line to obtain the crack line length and scratch line length; Arrange the crack line lengths and scratch line lengths in descending order according to their numerical values, extract the maximum crack line length and the maximum scratch line length, sum and average the maximum crack line length and obtain the crack line mean value; obtain the crack line mean value of each judgment sub-region in turn, and extract the largest crack line mean value, which is recorded as the crack line mean extreme value; Obtaining patterns formed by each crack line and scratch line and their corresponding crack characteristic profiles and scratch characteristic profiles, and recording them as crack surfaces and scratch surfaces respectively; Obtain the size and number of overlapping areas corresponding to the overlap between the crack surface and the scratch surface in each judgment sub-region in turn; extract the maximum overlapping area in each judgment sub-region in turn, collect the maximum overlapping areas from all judgment sub-regions together, and sort them in descending order according to the size of the overlapping areas, extract the maximum overlapping area among them, and record it as the maximum overlapping area value; According to the number of overlapping areas in each judgment sub-region, the two regions with the largest number of overlapping areas are extracted, and the shortest distance between the two regions is calculated and recorded as the span value; The crack coefficient is obtained by comprehensively processing the crack line extreme value, overlapping maximum area value and span value.
[0011] Preferably, the process of obtaining the sudden change coefficient includes: Calculate the gradient value of each pixel in the axonal image based on the gradient operator; A gradient threshold is preset, and the gradient value of each pixel in the axonal tube image is compared with the gradient threshold, and the gradient value greater than the preset gradient threshold is recorded as the gradient fluctuation value; Obtain the contour patterns formed by all gradient fluctuation values in the axle tube image, calculate the area of all contour patterns and divide them by the corresponding axle tube area to obtain the fluctuation degree; Calculate the difference between each gradient fluctuation value and the gradient threshold, and take the absolute value to obtain the gradient fluctuation difference; preset the abnormal value range of the gradient fluctuation difference, and compare the gradient fluctuation difference corresponding to each contour pattern with the abnormal value range of the gradient fluctuation difference in turn, and mark the gradient fluctuation difference within the abnormal value range of the gradient fluctuation difference to obtain the marked line segment corresponding to each contour pattern; Take the first and last ends of the marked line segments of the outline pattern as endpoints, and connect the endpoints of each marked line segment with a straight line in a clockwise order to obtain a closed figure formed by the marked line segments. Calculate the area of the closed figure and divide it by the area of the figure formed by the corresponding outline to obtain the proportion; Obtain the proportions corresponding to each contour in turn, sort the obtained proportions in descending order according to their numerical values, and extract the maximum proportion, which is recorded as the limit proportion; The sudden drop coefficient is obtained by comprehensively processing the fluctuation degree and the limit proportion.
[0012] Preferably, the process of obtaining the roughness coefficient includes: Obtain the shaft tube image, and extract the contour curve of the opening in the shaft tube image based on the image processing library, and record it as the comparison contour; A standard curve for the profile of the shaft tube opening is preset, and the comparison profile is overlapped with the standard curve. The non-overlapping part is recorded as a deviation curve. The deviation curves in each comparison profile are obtained in sequence, and the lengths of all deviation curves are calculated and summed to obtain the total deviation length. The perimeters of all standard curves of the opening profiles are calculated and summed up to obtain the standard total perimeter. The total deviation length is divided by the standard total perimeter to obtain the deviation degree. Arrange the deviation curves corresponding to each opening in descending order by length, and determine the opening positions corresponding to the three largest deviation curve lengths. Use the resulting circles of the three openings as endpoints, and connect the three endpoints with straight lines to form a complete triangle. Calculate the area of the triangle and record it as the triangle quantization value. The roughness coefficient is obtained by weighting the deviation and the triangular quantization value.
[0013] Preferably, the defect assessment coefficient is obtained by comprehensively processing the crack coefficient, the depression coefficient, and the roughness coefficient, specifically including: After normalizing the crack coefficient, sag coefficient, and roughness coefficient, the product of the crack coefficient and the sag coefficient is used as the side of an equilateral triangle to construct an equilateral triangle. The roughness coefficient is used as the height of the equilateral triangle to construct a triangular pyramid model. The volume of the triangular pyramid model is calculated and recorded as the defect assessment coefficient.
[0014] Preferably, the value ranges of the three groups of thresholds are preset, and the value range of each group of thresholds corresponds to a defect level. The defect assessment coefficient is matched with the value ranges of the three groups of thresholds to obtain the defect level corresponding to the defect assessment coefficient, where the defect level includes primary defect, secondary defect and tertiary defect.
[0015] Preferably, when the defect level corresponding to the evaluation coefficient is level one: it is considered qualified and included in the normal production process; When the defect level corresponding to the evaluation coefficient is level 2: the product is not directly deemed qualified, and whether to repair or release the product must be determined based on the defect location and functional impact; When the defect level corresponding to the evaluation coefficient is level three: it is judged as unqualified and must be processed compulsorily to prevent it from flowing into the next process or being delivered.
[0016] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. The present invention constructs three core indicators, namely the crack coefficient, the sag coefficient and the roughness coefficient, to three-dimensionally characterize the shaft tube defects from the three dimensions of "linear defect extensibility", "surface undulation severity" and "opening processing accuracy". The crack coefficient comprehensively considers the maximum line length, overlapping area and spatial distribution of cracks / scratches, and can accurately capture the expansion trend and aggregation risk of defects. The sag coefficient effectively distinguishes between gentle deformation and sharp defects through the global proportion of gradient fluctuations and the local abnormal concentration. The roughness coefficient quantifies the stability of the processing technology through the spatial correlation between the overall proportion of the opening profile deviation and the local extreme deviation.
[0017] 2. The present invention deeply links computer vision inspection results with the production process to form an automated closed loop of "inspection, evaluation, graded processing, and root cause tracing." For first-level defects, by establishing quality archives to statistically analyze the defect distribution patterns, early warning of equipment wear or process fluctuations can be provided. The differentiated processing strategy for second-level defects reduces repair costs while ensuring product performance. The mandatory scrapping and full-process traceability mechanism for third-level defects can quickly identify systemic risk sources and prevent unqualified products from entering the market. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Further details, features and advantages of the present application are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which: Figure 1 Flowchart of the present invention. DETAILED DESCRIPTION
[0019] Several embodiments of the present application will be described in more detail below with reference to the accompanying drawings so that those skilled in the art can implement the present application. The present application can be embodied in many different forms and for many different purposes and should not be limited to the embodiments described herein. These embodiments are provided to make the present application comprehensive and complete and to fully convey the scope of the present application to those skilled in the art. The embodiments do not limit the present application.
[0020] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. It will be further understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the relevant art and / or the context of this specification, and will not be interpreted in an idealized or overly formal sense unless expressly defined as such herein.
[0021] See also Figure 1 As shown, the present invention provides a technical solution: A method for identifying axle defects of a semi-trailer based on computer vision, comprising: Image data acquisition: obtain the axle tube image information of the axle; Specifically include: The shaft tube is photographed in real time from multiple angles using fixed-position cameras and light sources, including: The camera is an industrial camera with a frame rate that matches the production line speed and supports line scan or area scan; Light sources include: backlighting and multi-angle ring light sources; Backlighting is used to highlight contours and detect dimensional deviations and holes; Multi-angle ring light source is used to eliminate shadows and enhance the contrast of surface cracks and scratches; Feature extraction: After pre-processing the image information of the shaft tube, the crack and scratch features, convex and concave features, and hole and burr features of the shaft tube are extracted; Convert the color image into a grayscale image to eliminate color interference; then use Gaussian filtering or median filtering to remove random noise and salt and pepper noise and smooth the image; Use histogram equalization or adaptive histogram equalization to improve image contrast and enhance the difference between defects and background; sharpen images through Laplacian operator or high-pass filtering to highlight edge details; Finally, illumination compensation, such as background subtraction or local normalization, is performed to eliminate the effects of uneven illumination, ensuring that cracks, scratches, bumps, depressions, holes, and burrs on the central axis tube are clearly visible in the image while maintaining the consistency of the overall image quality. Feature analysis: After analyzing the crack and scratch features, protrusion and depression features, and hole burr features, the crack coefficient, protrusion coefficient, and roughness coefficient are obtained; The process of obtaining the crack coefficient includes: According to the preset size and area, the axle tube is evenly divided into regions to obtain various judgment sub-regions; Obtain the crack features and scratch features of each judgment sub-region, and connect the beginning and end points of each crack and scratch with a straight line to obtain the crack line length and scratch line length; Arrange the crack line lengths and scratch line lengths in descending order according to their numerical values, extract the maximum crack line length and the maximum scratch line length, sum and average the maximum crack line length and obtain the crack line mean value; obtain the crack line mean value of each judgment sub-region in turn, and extract the largest crack line mean value, which is recorded as the crack line mean extreme value; Obtaining patterns formed by each crack line and scratch line and their corresponding crack characteristic profiles and scratch characteristic profiles, and recording them as crack surfaces and scratch surfaces respectively; Obtain the size and number of overlapping areas corresponding to the overlap between the crack surface and the scratch surface in each judgment sub-region in turn; extract the maximum overlapping area in each judgment sub-region in turn, collect the maximum overlapping areas from all judgment sub-regions together, and sort them in descending order according to the size of the overlapping areas, extract the maximum overlapping area among them, and record it as the maximum overlapping area value; According to the number of overlapping areas in each judgment sub-region, the two regions with the largest number of overlapping areas are extracted, and the shortest distance between the two regions is calculated and recorded as the span value; The shortest distance between two regions is the line connecting the centers of the two regions; The crack coefficient is obtained by comprehensively processing the crack line extreme value, overlap maximum value and span value; The crack line extreme value, overlap maximum value and span value are marked as 、 、 The subsequent entry formula: , and the crack coefficient is obtained ; in 、 、 They are the reference crack line extreme value, the maximum allowable overlap maximum surface value and the reference span value respectively; 、 、 are the weight factors corresponding to the average extreme value of the crack line, the maximum overlap value and the span value respectively; The process of obtaining the subsidence coefficient includes: The gradient value of each pixel in the axle tube image is calculated based on the gradient operator. The calculation process is directly referenced from the existing technology and will not be described in detail here; A gradient threshold is preset, and the gradient value of each pixel in the axonal tube image is compared with the gradient threshold, and the gradient value greater than the preset gradient threshold is recorded as the gradient fluctuation value; Obtain the contour patterns formed by all gradient fluctuation values in the axle tube image, calculate the area of all contour patterns and divide them by the corresponding axle tube area to obtain the fluctuation degree; Calculate the difference between each gradient fluctuation value and the gradient threshold, and take the absolute value to obtain the gradient fluctuation difference; preset the abnormal value range of the gradient fluctuation difference, and compare the gradient fluctuation difference corresponding to each contour pattern with the abnormal value range of the gradient fluctuation difference in turn, and mark the gradient fluctuation difference within the abnormal value range of the gradient fluctuation difference to obtain the marked line segment corresponding to each contour pattern; The first and last ends of the marked segments of the outline pattern are used as endpoints, and the endpoints of each marked segment are connected with straight lines in a clockwise or counterclockwise order to obtain a closed figure formed by the marked segments. The area of the closed figure is calculated and divided by the area of the figure formed by the corresponding outline to obtain the proportion; Obtain the proportions corresponding to each contour in turn, sort the obtained proportions in descending order according to their numerical values, and extract the maximum proportion, which is recorded as the limit proportion; The sudden drop coefficient is obtained by comprehensively processing the fluctuation degree and the limit proportion; After normalizing the fluctuation degree and the limit proportion, the fluctuation degree and the limit proportion are used as the major and minor axes of the ellipse respectively, and an ellipse model is constructed. The area of the ellipse model is calculated as the indentation coefficient. The gradient operator is used to filter out the gradient fluctuation value in the tube image. The fluctuation value is used to quantify the overall coverage of surface defects, and the extreme ratio is used to highlight the concentration of local areas of severe mutation. These two methods comprehensively capture the convex and concave features of the tube surface from both global and local dimensions. Whether it is comprehensive processing to obtain the concave coefficient or constructing an elliptical model for quantification, complex surface defect characteristics can be converted into a single value, achieving quantitative and multi-dimensional assessment of tube surface defects, effectively avoiding the one-sidedness of a single indicator. The results are objective and comparable, and can accurately locate high-risk defect areas, providing a scientific basis for tube quality control and process optimization. The process of obtaining the roughness coefficient includes: Obtain the shaft tube image, and extract the contour curve of the opening in the shaft tube image based on the image processing library, and record it as the comparison contour; A standard curve for the profile of the shaft tube opening is preset, and the comparison profile is overlapped with the standard curve. The non-overlapping part is recorded as a deviation curve. The deviation curves in each comparison profile are obtained in sequence, and the lengths of all deviation curves are calculated and summed to obtain the total deviation length. The perimeters of all standard curves of the opening profiles are calculated and summed up to obtain the standard total perimeter. The total deviation length is divided by the standard total perimeter to obtain the deviation degree. Arrange the deviation curves corresponding to each opening in descending order by length, and determine the opening positions corresponding to the three largest deviation curve lengths. Use the resulting circles of the three openings as endpoints, and connect the three endpoints with straight lines to form a complete triangle. Calculate the area of the triangle and record it as the triangle quantization value. The roughness coefficient is obtained by weighting the deviation and triangular quantization value; Preset the weight factors of the deviation and triangular quantization value, multiply the deviation and triangular quantization value with their corresponding weight factors, and sum them to obtain the roughness coefficient; By extracting the comparison profile of the shaft tube opening and comparing it with the standard curve, the deviation of the overall profile is quantified using deviation metric, and the spatial distribution of local significant defects is located and characterized using triangulated quantization values. Finally, the roughness coefficient is obtained through weighted calculation, and the complex surface roughness is converted into a single value. This achieves a quantitative and structured assessment of the roughness of the shaft tube opening. It not only avoids the subjectivity of manual visual inspection, but also takes into account both overall and local defects, and comprehensively covers the different manifestations of roughness. The numerical results can be directly connected to production standards, providing an objective basis for processing technology optimization and quality grading. Comprehensive processing: The crack coefficient, depression coefficient and roughness coefficient are processed comprehensively to obtain the defect assessment coefficient; Specifically include: After normalizing the crack coefficient, the protrusion coefficient, and the roughness coefficient, the product of the crack coefficient and the protrusion coefficient is used as the side of an equilateral triangle to construct an equilateral triangle. The roughness coefficient is used as the height of the equilateral triangle to construct a triangular pyramid model. The volume of the triangular pyramid model is calculated and recorded as the defect assessment coefficient. Defect level assessment: Match the corresponding defect level based on the defect assessment coefficient and perform corresponding processing; Three groups of threshold value ranges are preset, and each threshold value range corresponds to a defect level. The defect assessment coefficient is matched with the three threshold value ranges to obtain the defect level corresponding to the defect assessment coefficient. The defect levels include primary defects, secondary defects, and tertiary defects. The defect level is proportional to the defect. When the defect level corresponding to the evaluation coefficient is level 1, it is considered qualified and incorporated into the normal production process. The defect evaluation coefficient and corresponding location are recorded and a quality file is established for traceability and process stability analysis. The frequency of level 1 defects is regularly counted. If the frequency increases abnormally, potential systemic risks such as equipment accuracy and raw materials (such as tool wear and temperature fluctuations) need to be investigated to prevent defects from escalating. When the defect level corresponding to the evaluation coefficient is Level 2: the product is not directly deemed qualified. The decision on whether to repair or release the product must be made based on the defect location and functional impact. If the defect is located on a non-critical mating surface (such as a non-load-bearing area or non-sealing surface) and mechanical simulation or actual measurement verify that it does not affect performance, the defect level can be relaxed to qualified, but the product must be marked as "restricted use" (e.g., only for specific working conditions). If the defect is located in a critical area (such as a bearing mounting surface or fluid channel), the product must be repaired. When the defect level corresponding to the evaluation coefficient is level three: it is judged as unqualified and must be processed compulsorily to prevent it from flowing into the next process or delivery; If the defect cannot be eliminated by repair (such as through cracks, large-area dimensional deviations), or the cost of repair exceeds the cost of manufacturing a new product, it should be directly scrapped and stored separately to avoid confusion; Conduct comprehensive failure analysis on scrapped parts (such as microscopic inspection and processing parameter backtracking) to identify the root causes (such as equipment failure, operational errors, and design defects); trace other shaft tubes produced in the same batch and process, expand the inspection scope (such as 100% full inspection), and avoid the outflow of batches of defective products.
[0022] The above formulas are obtained by collecting a large amount of data and performing software simulation, and a formula close to the actual value is selected. The influencing weight factors and specific coefficient values in the formula are set by technical personnel in this field according to actual conditions, and can be adjusted and modified later.
[0023] The above description of the embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for identifying axle defects of a semi-trailer based on computer vision, characterized in that: include: Image data acquisition: obtain the axle tube image information of the axle; Feature extraction: After pre-processing the image information of the shaft tube, the crack and scratch features, convex and concave features, and hole and burr features of the shaft tube are extracted; feature Analysis: The crack and scratch characteristics, protrusion and depression characteristics, and hole burr characteristics are analyzed to obtain the crack coefficient, protrusion coefficient, and roughness coefficient; Comprehensive processing: The crack coefficient, depression coefficient and roughness coefficient are processed comprehensively to obtain the defect assessment coefficient; Defect level assessment: Match the corresponding defect level based on the defect assessment coefficient and perform corresponding processing.
2. The method for identifying axle defects of a semitrailer based on computer vision according to claim 1, characterized in that: Image data acquisition, specifically including: The shaft tube is photographed in real time from multiple angles using fixed-position cameras and light sources, including: The camera is an industrial camera with a frame rate that matches the production line speed and supports line scan; The light source includes: backlight and multi-angle ring light source.
3. The method for identifying axle defects of a semi-trailer based on computer vision according to claim 2, characterized in that: The process of preprocessing the image information of the shaft tube specifically includes: Convert the color image into a grayscale image to eliminate color interference; then use Gaussian filtering or median filtering to remove random noise and salt and pepper noise and smooth the image; Use histogram equalization or adaptive histogram equalization to improve image contrast and enhance the difference between defects and background; sharpen images through Laplacian operator or high-pass filtering to highlight edge details; Finally, illumination compensation is performed to eliminate the effects of uneven illumination, ensuring that cracks, scratches, bumps, depressions, holes, and burr defects on the central axis tube of the image are clearly discernible while maintaining the consistency of the overall image quality.
4. The method for identifying axle defects of a semitrailer based on computer vision according to claim 3, characterized in that: The process of obtaining the crack coefficient includes: According to the preset size and area, the axle tube is evenly divided into regions to obtain various judgment sub-regions; Obtain the crack features and scratch features of each judgment sub-region, and connect the beginning and end points of each crack and scratch with a straight line to obtain the crack line length and scratch line length; Arrange the crack line lengths and scratch line lengths in descending order according to their numerical values, extract the maximum crack line length and the maximum scratch line length, sum and average the maximum crack line length and obtain the crack line mean value; obtain the crack line mean value of each judgment sub-region in turn, and extract the largest crack line mean value, which is recorded as the crack line mean extreme value; Obtaining patterns formed by each crack line and scratch line and their corresponding crack characteristic profiles and scratch characteristic profiles, and recording them as crack surfaces and scratch surfaces respectively; Obtain the size and number of overlapping areas corresponding to the overlap between the crack surface and the scratch surface in each judgment sub-region in turn; extract the maximum overlapping area in each judgment sub-region in turn, collect the maximum overlapping areas from all judgment sub-regions together, and sort them in descending order according to the size of the overlapping areas, extract the maximum overlapping area among them, and record it as the maximum overlapping area value; According to the number of overlapping areas in each judgment sub-region, the two regions with the largest number of overlapping areas are extracted, and the shortest distance between the two regions is calculated and recorded as the span value; The crack coefficient is obtained by comprehensively processing the crack line extreme value, overlapping maximum area value and span value.
5. The method for identifying axle defects of a semi-trailer based on computer vision according to claim 4, characterized in that: The process of obtaining the subsidence coefficient includes: Calculate the gradient value of each pixel in the axonal image based on the gradient operator; A gradient threshold is preset, and the gradient value of each pixel in the axonal tube image is compared with the gradient threshold, and the gradient value greater than the preset gradient threshold is recorded as the gradient fluctuation value; Obtain the contour patterns formed by all gradient fluctuation values in the axle tube image, calculate the area of all contour patterns and divide them by the corresponding axle tube area to obtain the fluctuation degree; Calculate the difference between each gradient fluctuation value and the gradient threshold, and take the absolute value to obtain the gradient fluctuation difference; preset the abnormal value range of the gradient fluctuation difference, and compare the gradient fluctuation difference corresponding to each contour pattern with the abnormal value range of the gradient fluctuation difference in turn, and mark the gradient fluctuation difference within the abnormal value range of the gradient fluctuation difference to obtain the marked line segment corresponding to each contour pattern; Take the first and last ends of the marked line segments of the outline pattern as endpoints, and connect the endpoints of each marked line segment with a straight line in a clockwise order to obtain a closed figure formed by the marked line segments. Calculate the area of the closed figure and divide it by the area of the figure formed by the corresponding outline to obtain the proportion; Obtain the proportions corresponding to each contour in turn, sort the obtained proportions in descending order according to their numerical values, and extract the maximum proportion, which is recorded as the limit proportion; The sudden drop coefficient is obtained by comprehensively processing the fluctuation degree and the limit proportion.
6. The method for identifying axle defects of a semitrailer based on computer vision according to claim 5, characterized in that: The process of obtaining the roughness coefficient includes: Obtain the shaft tube image, and extract the contour curve of the opening in the shaft tube image based on the image processing library, and record it as the comparison contour; A standard curve for the profile of the shaft tube opening is preset, and the comparison profile is overlapped with the standard curve. The non-overlapping part is recorded as a deviation curve. The deviation curves in each comparison profile are obtained in sequence, and the lengths of all deviation curves are calculated and summed to obtain the total deviation length. The perimeters of all standard curves of the opening profiles are calculated and summed up to obtain the standard total perimeter. The total deviation length is divided by the standard total perimeter to obtain the deviation degree. Arrange the deviation curves corresponding to each opening in descending order by length, and determine the opening positions corresponding to the three largest deviation curve lengths. Use the resulting circles of the three openings as endpoints, and connect the three endpoints with straight lines to form a complete triangle. Calculate the area of the triangle and record it as the triangle quantization value. The roughness coefficient is obtained by weighting the deviation and the triangular quantization value.
7. The method for identifying axle defects of a semitrailer based on computer vision according to claim 6, characterized in that: The defect assessment coefficient is obtained by comprehensively processing the crack coefficient, the depression coefficient, and the roughness coefficient, which specifically includes: After normalizing the crack coefficient, sag coefficient, and roughness coefficient, the product of the crack coefficient and the sag coefficient is used as the side of an equilateral triangle to construct an equilateral triangle. The roughness coefficient is used as the height of the equilateral triangle to construct a triangular pyramid model. The volume of the triangular pyramid model is calculated and recorded as the defect assessment coefficient.
8. The method for identifying axle defects of a semitrailer based on computer vision according to claim 7, characterized in that: Three groups of threshold value ranges are preset, and the value range of each group of threshold values corresponds to a defect level. The defect assessment coefficient is matched with the value range of the three groups of threshold values to obtain the defect level corresponding to the defect assessment coefficient, where the defect level includes primary defect, secondary defect and tertiary defect.
9. The method for identifying axle defects of a semitrailer based on computer vision according to claim 8, characterized in that: When the defect level corresponding to the evaluation coefficient is level 1: it is considered qualified and included in the normal production process; When the defect level corresponding to the evaluation coefficient is level 2: the product is not directly deemed qualified, and whether to repair or release the product must be determined based on the defect location and functional impact; When the defect level corresponding to the evaluation coefficient is level three: it is judged as unqualified and must be processed compulsorily to prevent it from flowing into the next process or being delivered.