Sole wear degree detection system based on image recognition

Abstract

The invention discloses a sole wear degree detection system based on image recognition, and relates to the technical field of image recognition, the system comprises an image acquisition module, an image processing module, a wear degree analysis module, a database module and a decision output module; the image acquisition module is used for acquiring a multi-angle high-definition image of the shoe sole; the image processing module is used for carrying out impurity removal processing on the multi-angle high-definition image; the database module is used for storing data of different shoe types; the wear degree analysis module is used for distinguishing normal wear from abnormal wear according to the database module and the three-dimensional distribution data; and the decision output module is used for outputting a quality guarantee judgment conclusion and a maintenance scheme. According to the shoe sole wear degree detection system based on image recognition, natural wear and abnormal wear are accurately distinguished through three-dimensional shape reconstruction and a deep learning model.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and particularly relates to a sole wear degree detection system based on image recognition. Background Art

[0002] For existing work shoes and insulating shoes, the soles have various patterns with different sizes and depths. During use, some smaller and shallower patterns will wear out earlier. Therefore, a sole wear degree detection system is adopted to detect the wear state of the soles.

[0003] Existing sole wear degree detection mainly relies on manual visual inspection or simple thickness measurement. The thickness of a single point in a local area of the sole is compared by a caliper or a handheld thickness gauge to judge the wear degree. Some technologies use two-dimensional image scanning equipment to obtain the planar image of the sole and identify the worn area based on the change of gray value, but the three-dimensional topographic difference cannot be analyzed.

[0004] Existing sole wear degree detection technologies have the following defects: 1. Existing technologies rely on manual visual inspection to distinguish natural wear from artificial damage, which is easily affected by the experience of quality inspectors. The recognition accuracy of malicious complaints is relatively low. There is a lack of correlation analysis between wearing habits and wear patterns, and it is difficult to identify malicious wear caused by short-term abnormal use. It is impossible to effectively detect sole forgery and fraud, and there is a risk of fraudulent claims; 2. Existing technologies only compare local thickness data and cannot comprehensively reflect the three-dimensional topographic changes of the sole, resulting in inaccurate formulation of maintenance plans.

[0005] Therefore, we propose a sole wear degree detection system based on image recognition. Summary of the Invention

[0006] The main purpose of the present invention is to provide a sole wear degree detection system based on image recognition, through which to solve the problems in the prior art that rely on manual visual inspection to distinguish natural wear from artificial damage, are easily affected by the experience of quality inspectors, have a relatively low recognition accuracy of malicious complaints, lack the correlation analysis between wearing habits and wear patterns, are difficult to identify malicious wear caused by short-term abnormal use, cannot effectively detect sole forgery and fraud, and there is a risk of fraudulent claims; and existing technologies only compare local thickness data and cannot comprehensively reflect the three-dimensional topographic changes of the sole, resulting in inaccurate formulation of maintenance plans.

[0007] To achieve the above object, the present invention provides the following technical solutions: A sole wear degree detection system based on image recognition, comprising: an image acquisition module, an image processing module, a wear degree analysis module, a database module, and a decision output module; The image acquisition module is used to acquire multi-angle high-definition images of the sole; An image processing module, which is used to remove impurities from multi-angle high-definition images, generate a three-dimensional image model of the sole according to the multi-angle high-definition images through a preset three-dimensional reconstruction unit, and analyze the three-dimensional image model and the preset standard three-dimensional morphology features of the sole through a preset processing unit to obtain the three-dimensional distribution data of the worn area of the sole; A database module, which is used to store the standard three-dimensional morphology features of soles of different shoe models, user wearing history data, a malicious wear feature library, and standard wear data; A wear degree analysis module, which performs the following determinations based on a pre-trained deep learning model: distinguishing normal wear and abnormal wear according to the database module and the three-dimensional distribution data, where the abnormal wear includes non-uniform deformation caused by sole quality problems and regular geometric wear caused by user malicious damage; and generating a quantitative wear score according to the normal wear and the abnormal wear; A decision output module, which is used to combine the quantitative wear score and a preset threshold to output a quality assurance determination conclusion and a repair plan.

[0008] Preferably, communication connections are established between the image acquisition module, the image processing module, the wear degree analysis module, the database module, and the decision output module; The image acquisition module includes a rotatable stage, an annular LED light source array, and a high-resolution industrial camera.

[0009] Preferably, the steps of removing impurities from the multi-angle high-definition images in the image processing module specifically include: S1. Calibrate the positions of the multi-angle high-definition images of the sole, eliminate the image offset caused by the shooting angle difference, separate the sole body and the shooting background in the multi-angle high-definition images of the sole through color recognition and contour extraction technology, delete the interference pixels of the rotatable stage, shadows, and non-sole areas, and retain multiple first complete sole images; S2. For the bright spots or specular effects generated by light reflection in the multiple first complete sole images, dynamically adjust the parameters of the annular LED light source array and the image brightness balance, eliminate the interference of the high-light area on the texture details, and automatically identify the missing parts of the multiple first complete sole images caused by sole stains or shooting angle limitations through a preset repair unit, predict and fill the image content of the missing area through the texture features of adjacent areas, and generate multiple second complete sole images; S3. Reduce the random noise in the multiple second complete sole images through a multi-level filtering method, smooth the granular noise generated by environmental interference, and strengthen the contour features of the sole pattern edges and worn areas, highlighting the transition difference between the worn boundary and the normal area, and completing the impurity removal process of the multi-angle high-definition images of the sole.

[0010] Preferably, the three-dimensional distribution data of the sole wear area in the image processing module is obtained by the following steps: Q1. The three-dimensional reconstruction unit uses a plurality of the second complete sole images that have undergone impurity removal processing to extract the spatial position information of the sole from different perspectives, calculates and generates a set of three-dimensional spatial data points on the sole surface through geometric relationships, and connects the spatial relationships between adjacent three-dimensional spatial data points according to the distribution law of the set of three-dimensional spatial data points to form a continuous three-dimensional image model of the sole; Q2. The processing unit performs mirror processing on the three-dimensional image model and the standard three-dimensional shape features of the sole, and presets a key point alignment method to eliminate the spatial deviation caused by the shooting angle. A preset reference datum plane is marked in the middle of the three-dimensional image model and the standard three-dimensional shape features of the sole, and the vertical distances between the three-dimensional image model and the reference datum plane, and between the standard three-dimensional shape features of the sole and the reference datum plane are calculated point by point to generate a first set of depth change values and a second set of depth change values; Q3. Compare the first set of depth change values with the second set of depth change values, analyze the values in the first set of depth change values that exceed or are less than the values in the second set of depth change values, generate a first three-dimensional wear distribution map of the depth difference, generate a second three-dimensional wear distribution map including the position and the area ratio according to the first three-dimensional wear distribution map, and mark the wear parameters of the key areas of the front sole, the heel and the arch of the sole in the first three-dimensional wear distribution map and the second three-dimensional wear distribution map to obtain the three-dimensional distribution data of the sole wear area.

[0011] Preferably, the specific steps of generating the second three-dimensional wear distribution map including the position and the area ratio according to the first three-dimensional wear distribution map in step Q3 are as follows: Q31. Based on the depth difference data of the first three-dimensional wear distribution map, delimit the areas on the surface of the three-dimensional image model where the depth change exceeds the preset threshold, mark them as wear candidate areas, and record the central coordinates and boundary ranges of each wear candidate area; Q32. For each wear candidate area, calculate the surface area of the sole covered by its boundary range, convert it into a ratio with the overall surface area of the sole, and obtain the area ratio value of each wear candidate area; Q33. According to the combined result of the depth difference data of the first three-dimensional wear distribution map and the area ratio value, divide the wear candidate areas into three wear levels: mild, moderate and severe; Q34. In the surface coordinate system of the three-dimensional image model, taking the central coordinates of the wear candidate area as the anchor point, a circular coverage area is generated, the radius of which is proportional to the area ratio value, and the color depth of the circular coverage area is positively correlated with the wear level; Q35. Mark the area ratio values of each wear candidate area at the center of the corresponding circular coverage area, and connect adjacent wear candidate areas through connection lines to form a second three-dimensional wear distribution map showing the wear distribution density and severity.

[0012] Preferably, the specific steps for distinguishing normal wear and abnormal wear in the wear degree analysis module are as follows: H1. Extract the position distribution, shape regularity, and depth difference data of each wear candidate area from the second three-dimensional wear distribution map, and count the area ratio and depth extreme values of the key areas of the front sole, heel, and arch of the shoe; H2. Compare the wear characteristics of the extracted wear candidate areas with the standard wear data in the database module, and analyze whether the wear candidate areas conform to the distribution law of human gait mechanics. If the wear is concentrated in common stress areas and the shape is naturally diffused, it is marked as normal wear; H3. When at least one of the following situations is detected, it is determined as abnormal wear: a. The wear candidate area has regular geometric features and uniform depth changes, conforming to the tool damage characteristics; b. There is an isolated wear area with a sudden depth change in the non-stressed area of the sole; c. The total area ratio of multiple wear areas on the same sole exceeds the normal threshold corresponding to the wearing duration; d. The texture similarity between the wear candidate area and the artificial damage cases in the malicious wear feature library exceeds 90%.

[0013] Preferably, the generation steps of the quantitative wear score in the wear degree analysis module are as follows: F1. According to the position, area ratio, and depth difference data of the wear candidate areas marked in the second three-dimensional wear distribution map, as well as the normal wear and abnormal wear, extract the cumulative area ratio, maximum depth value, number of abnormal wear areas, and distribution density parameters of each wear candidate area, and combine the standard wear data of the same shoe type in the database module to assign dynamic weights to the wear parameters of different wear candidate areas. Among them, the area ratio weights of the front sole and heel areas of the shoe are higher than those of the arch area, and the weight of the abnormal wear area is 3 times that of the normal wear area; F2. Call the user's wearing history data in the database module, and adjust the benchmark threshold of the wear score according to the actual wearing duration and usage scenario; F3. If there is an abnormal wear or a wear candidate area with normal wear, an additional scoring deduction value is added according to the matching degree between its regular geometric features and the malicious wear feature library. The deduction value is proportional to the area ratio of the abnormal wear area and inversely proportional to the area ratio of the normal wear area. F3. Superimpose the wear parameters, the reference threshold, and the deduction value with the dynamically assigned weights to generate a quantitative wear score ranging from 0 to 100 points, and divide it into four grades: more than 80 points is slight wear, 60 - 80 points is moderate wear, 30 - 60 points is severe wear, and less than 30 points is serious damage.

[0014] Preferably, the specific operation process of the decision output module is as follows: C1. According to the grade corresponding to the quantitative wear score, match the preset repair plan library and extract the repair costs corresponding to each grade. C2. According to the type of abnormal wear, obtain the following quality assurance determination conclusions: If the abnormal wear is caused by non - uniform deformation due to sole quality problems, output the conclusion of "meeting the quality assurance scope"; if the abnormal wear conforms to the regular geometric wear characteristics caused by user malicious damage, output the conclusion of "artificial damage"; if the quantitative wear score of the normal wear area is lower than the preset threshold but does not trigger abnormal determination, output the conclusion of "normal loss". C3. Generate corresponding repair steps according to the repair cost and the quality assurance determination conclusion, integrate them into a visual report, mark the repair priority and cost details of each wear area, and output them to a preset terminal.

[0015] Compared with the prior art, the present invention has the following beneficial effects: First, in the present invention, through three - dimensional topography reconstruction and deep - learning models, the system accurately distinguishes natural wear and abnormal wear. Combining the texture matching technology of the malicious wear feature library, it effectively identifies malicious tampering behaviors caused by artificial damage or short - term abnormal use. Compared with manual visual inspection, the detection accuracy is increased to more than 90%, significantly reducing the fraud risk in insurance claims. At the same time, it supports sole quality traceability and provides objective evidence for merchants' malicious complaints. Second, in the present invention, by adopting multi - angle three - dimensional reconstruction technology, the depth difference between the actual sole and the standard model is compared point by point through the reference plane, accurately capturing the local thickness change and the overall deformation trend, avoiding the one - sidedness of traditional single - thickness detection. Combining multi - parameter dynamic weight evaluation such as the area ratio of the wear area, depth extreme values, and distribution density, a quantitative score and a three - dimensional distribution map are generated, providing accurate data support for the repair plan, improving the repair efficiency and reducing the return rate. III. In the present invention, a maintenance plan and a quality assurance conclusion are automatically matched based on a quantitative score, and a visual report is output, reducing the manual intervention cost and increasing the quality inspection efficiency by 80%. It is applicable to the large-scale inspection requirements of shoe enterprises and insurance institutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is the system flowchart of the present invention; Figure 2 is the flowchart of removing impurities from multi-level high-definition images of the present invention; Figure 3 is the flowchart of generating three-dimensional distribution data of the present invention; Figure 4 is the flowchart of generating a quantitative wear score of the present invention.

[0017] In the figure: 1. Image acquisition module; 2. Image processing module; 3. Wear degree analysis module; 4. Database module; 5. Decision output module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] In order to make the technical means, creative features, achieved purposes and functions of the present invention easy to understand, the present invention will be further described below in conjunction with specific embodiments.

[0019] As Figures 1-4 shown, a sole wear degree detection system based on image recognition includes: an image acquisition module 1, an image processing module 2, a wear degree analysis module 3, a database module 4 and a decision output module 5; The image acquisition module 1 is used to obtain multi-angle high-definition images of the sole; the image acquisition module 1 includes a rotatable stage, an annular LED light source array and a high-resolution industrial camera. The image acquisition module 1 automatically adjusts the shoe body angle through the rotatable stage, cooperates with the annular LED light source to dynamically adjust the light to eliminate the reflection interference, and the high-resolution industrial camera synchronously takes multi-angle high-definition images. The stage rotates precisely to ensure 360-degree coverage, the light source intelligently adapts to the sole material to balance the brightness and color temperature, and the high-resolution industrial camera captures details at a resolution of 20 million pixels. The three work together to achieve non-blind spot and high-clear image acquisition, providing standardized data for subsequent 3D reconstruction, completing multi-angle shooting of a single shoe within 10 seconds, and being compatible with soles of different materials and colors; The image processing module 2 is used to remove impurities from the multi-angle high-definition images, generate a three-dimensional image model of the sole according to the multi-angle high-definition images through a preset three-dimensional reconstruction unit, and analyze the three-dimensional image model and a preset standard three-dimensional shape feature of the sole through a preset processing unit to obtain three-dimensional distribution data of the sole wear area; Among them, the specific steps for removing impurities from the multi-angle high-definition images in the image processing module 2 include: S1. Calibrate the positions of the multi-angle high-definition images of the shoe sole to eliminate image offsets caused by differences in shooting angles. Through color recognition and contour extraction techniques, separate the shoe sole body from the shooting background in the multi-angle high-definition images of the shoe sole, delete the interfering pixels in the rotatable stage, shadows, and non-shoe sole areas, and retain multiple first complete shoe sole images; The position calibration and background separation of the multi-angle shoe sole images are achieved through the following process: First, based on the rotation angle parameters of the rotatable stage, locate the preset marker points (such as the positioning marks on the edge of the stage) in each shoe sole image, and align the images taken at different angles to a unified spatial reference through a coordinate system conversion algorithm to eliminate the perspective offset caused by rotation. Subsequently, utilize the color gamut difference between the shoe sole and the background (such as the shoe sole is made of dark rubber and the stage is made of light matte material), extract the contour of the shoe sole body through HSV color space threshold segmentation, strengthen the shoe sole boundary by combining edge detection techniques, intelligently identify and eliminate interference areas such as the edge of the stage, device reflection, and projection shadows. Finally, through pixel-level mask operation, only retain the shoe sole body area to generate a pure image without background interference, ensuring the accuracy of subsequent 3D reconstruction; S2. For the bright spots or specular effects generated by light reflection in the multiple first complete shoe sole images, dynamically adjust the parameters of the annular LED light source array and the image brightness balance to eliminate the interference of the high-light area on the texture details, and automatically identify the missing parts of some areas in the multiple first complete shoe sole images caused by shoe sole stains or shooting angle limitations through a preset repair unit. Predict and fill the image content of the missing areas through the texture features of adjacent areas to generate multiple second complete shoe sole images; In step S2, the elimination of the interference of the high-light area on the texture details specifically includes: S21. Automatically identify the over-bright areas formed by reflection in the first complete shoe sole image, and judge the interference range according to the brightness and color characteristics; S22. Dynamically adjust the irradiation angle and intensity of the annular LED light source array, re-shoot or adjust the light conditions for the high-light area to reduce overexposure; S23. Combine the clear image taken after adjusting the light with the first complete shoe sole image, and cover the high-light part in the first complete shoe sole image with the details of the normal area to ensure the integrity of the texture; The steps for generating the second complete shoe sole image in step S2 are as follows: S24. Detect the areas where the patterns are broken and the colors are inconsistent in the first complete shoe sole image due to stains or occlusion, and mark the missing range; S25. Analyze the characteristics such as the pattern direction and particle density around the missing area, and speculate on the patterns and structures that should be in the missing part; S26. Generate filling content consistent with the surroundings according to the speculation result and smoothly connect the boundaries, so that the repaired area is naturally integrated with the first complete sole image to generate a second complete sole image; S3. Reduce random noise in multiple second complete sole images through a multi-level filtering method, smooth the granular noise caused by environmental interference, and enhance the contour features of the sole pattern edges and worn areas, highlighting the transition difference between the worn boundary and the normal area, and completing the noise removal process for the multi-angle high-definition images of the sole; Among them, the noise removal process for the multi-angle high-definition images of the sole in step S3 includes the following steps: S31. Random noise elimination: Perform preliminary filtering on the second complete sole image, reduce the brightness difference of isolated pixel points in the image layer by layer, reduce the fine noise caused by environmental interference, and at the same time retain the main structure of the sole pattern; S32. Granular noise smoothing: In the image after preliminary noise reduction, for the granular noise formed on the sole surface due to dust or uneven material, gradually reduce the brightness contrast of the granular area through local blurring and edge retention techniques to make the surface texture transition smooth; S33. Contour feature enhancement: Perform contrast adaptive adjustment on the processed image, enhance the brightness difference of the sole pattern edges, and highlight the transition line between the worn area and the normal area through edge detection technology, so that the worn contour is clearly visible, and finally output a high-precision noise-removed image.

[0020] Embodiment 1: Fix a pair of sports shoes on a rotatable loading platform. After starting the system, the rotatable loading platform automatically rotates 8 times at intervals of 45°. The annular LED light source array is dynamically adjusted to the warm light mode according to the sole material, and the high-resolution industrial camera synchronously takes 20 million pixel high-definition images. The image processing module aligns the images at each angle through coordinate system conversion, extracts the sole main body using HSV color space segmentation and clears the background interference; for the reflective area of the front sole, the system automatically reduces the brightness of the corresponding LED partition and takes a new picture to replace the high-light part of the original picture; then detects the missing pattern blocked by stains at the arch of the sole, generates filling content based on adjacent textures, and outputs a high-precision noise-removed image after multi-level noise reduction and contour enhancement.

[0021] Among them, the three-dimensional distribution data of the worn area of the sole in the image processing module 2 is obtained by the following steps: Q1. The three-dimensional reconstruction unit uses multiple second complete sole images that have undergone noise removal processing to extract the spatial position information of the sole from different perspectives, calculates and generates a set of three-dimensional spatial data points on the sole surface through geometric relationships, and connects the spatial relationships between its adjacent three-dimensional spatial data points according to the distribution law of the three-dimensional spatial data points to form a continuous three-dimensional image model of the sole; Among them, in step Q1, a three-dimensional spatial data point set of the sole surface is generated through geometric relationship calculations. The specific content includes: First, based on the position offsets of the same sole feature point (such as the pattern intersection point) in multiple second complete sole images under different perspectives, combined with the camera parameters (focal length, shooting angle), a spatial geometric model is constructed to calculate the three-dimensional coordinates of each feature point. Subsequently, based on the relative distance and angular relationship between adjacent feature points, the surface curvature of the unshot area is inferred to generate a dense data point set covering the entire sole. Finally, a continuous three-dimensional surface is formed through point cloud interpolation. Q2. The processing unit performs mirror processing on the three-dimensional image model and the standard three-dimensional shape features of the sole, and presets a key point alignment method to eliminate the spatial deviation caused by the shooting angle. A preset reference datum plane is marked in the middle of the three-dimensional image model and the standard three-dimensional shape features of the sole. The vertical distances between the three-dimensional image model and the reference datum plane, and between the standard three-dimensional shape features of the sole and the reference datum plane are calculated point by point to generate a first set of depth change values and a second set of depth change values. Among them, in step Q2, the preset key point alignment method eliminates the spatial deviation caused by the shooting angle. The steps are as follows: At least five groups of corresponding feature points are selected as alignment benchmarks at the forefoot, heel center, and pattern nodes of the sole of the three-dimensional image model and the standard three-dimensional shape features of the sole. The coordinate position of the three-dimensional image model is adjusted through rotation and translation operations to minimize the spatial position deviation between the two groups of feature points and eliminate the spatial offset caused by the shooting angle difference. Subsequently, based on the consistency of the surface curvature of the three-dimensional image model and the standard three-dimensional shape features of the sole, the alignment accuracy is further optimized until the overall shape matches. In step Q2, the vertical distances between the three-dimensional image model and the reference datum plane, and between the standard three-dimensional shape features of the sole and the reference datum plane are calculated point by point. The specific content includes: A horizontal reference datum plane is set between the three-dimensional image model and the standard three-dimensional shape features of the sole. The shortest vertical distance between each point on the surface of the three-dimensional image model and the datum plane, and the vertical distance between each point on the surface of the standard three-dimensional shape features of the sole and the datum plane are calculated respectively to generate two sets of depth values, which are used to quantify the deformation difference between the actual sole and the standard three-dimensional shape features of the sole. Q3. The first set of depth change values and the second set of depth change values are compared. The values in the first set of depth change values that exceed or are less than the values in the second set of depth change values are analyzed to generate a first three-dimensional wear distribution map of the depth difference. A second three-dimensional wear distribution map including the position and area ratio is generated based on the first three-dimensional wear distribution map. The wear parameters of the key areas of the forefoot, heel, and arch of the sole are marked in the first three-dimensional wear distribution map and the second three-dimensional wear distribution map to obtain the three-dimensional distribution data of the worn area of the sole.

[0022] Among them, the specific steps of generating the second three-dimensional wear distribution map including position and area ratio according to the first three-dimensional wear distribution map in step Q3 are as follows: Q31. Based on the depth difference data of the first three-dimensional wear distribution map, delimit the areas on the surface of the three-dimensional image model where the depth change exceeds the preset threshold, mark them as wear candidate areas, and record the central coordinates and boundary ranges of each wear candidate area; Q32. For each wear candidate area, calculate the sole surface area covered according to its boundary range, convert it into a ratio with the overall surface area of the sole, and obtain the area ratio value of each wear candidate area; Among them, the calculation method of the area ratio value of each wear candidate area is: divide the surface of the sole three-dimensional image model into tiny unit grids, calculate the actual area of each grid, accumulate the total area of all grids within the boundary range of the wear candidate area to obtain the surface area of the candidate area; then convert the sum of the areas of each candidate area into a percentage with the overall surface area of the sole (i.e., the sum of all grid areas) to obtain the area ratio value of each area; Q33. According to the combined results of the depth difference data and area ratio values of the first three-dimensional wear distribution map, divide the wear candidate areas into three wear levels: mild, moderate, and severe; Q34. In the surface coordinate system of the three-dimensional image model, with the central coordinates of the wear candidate area as the anchor point, generate a circular coverage area, the radius of which is proportional to the area ratio value, and the color depth of the circular coverage area is positively correlated with the wear level; Q35. Mark the area ratio value of each wear candidate area at the center of the corresponding circular coverage area, and connect adjacent wear candidate areas through connection lines to form the second three-dimensional wear distribution map showing the wear distribution density and severity.

[0023] Embodiment 2: After a pair of hiking shoes is photographed by the image acquisition module, the image processing module 2 executes the following process: Three-dimensional reconstruction: Extract the features of the sole arch groove, forefoot anti-slip pattern, and heel support point based on the multi-angle images after removing impurities, and generate a three-dimensional model containing 280,000 data points; Key point alignment: Select the center depression point of the arch, the corresponding pattern node of the first metatarsal bone of the forefoot, the center point of the heel, and 2 pattern intersection points as the alignment reference, and adjust the model position so that the overall coordinate deviation ≤ 0.25 mm; Depth comparison: Taking the plane 1.2 mm below the lowest point of the sole as the reference plane, it is detected that: the depth difference in the forefoot area is 1.1 mm (standard 0.6 mm); the difference in the heel area is 0.8 mm (standard 0.4 mm); the difference in the arch area is 0.4 mm (standard 0.2 mm); Wear distribution generation: In the forefoot, delimit 130 mm exceeding the threshold (0.8 mm) 2area, accounting for the total surface area of ​​the sole (1150mm 2 ) of 11.3%; the heel is defined as 50mm above the threshold (0.6mm) 2 Area, accounting for 4.3%; arch demarcation exceeds the threshold (0.3mm) 35mm 2 area, accounting for 3.0%; Output results: The wear coordinates, area proportion (11.3%, 4.3%, 3.0%) and depth difference data of the forefoot, heel and arch are marked in the 3D model.

[0024] Database module 4, which is used to store standard three-dimensional morphological features of the soles of different shoe types, user wearing history data, malicious wear feature library and standard wear data. Database module 4 adopts a sub-library and sub-table storage method: three-dimensional morphological features are stored in a dedicated model library in a point cloud encryption format; user wearing history data is stored in a relational database in time series; the malicious wear feature library and standard wear data are stored in a structured table in association with each other, and a feature vector index is established to support fast comparison and retrieval; Wear analysis module 3, wear analysis module 3 performs the following judgments based on the pre-trained deep learning model: according to database module 4 and three-dimensional distribution data, distinguish normal wear from abnormal wear, abnormal wear includes non-uniform deformation caused by sole quality problems and regular geometric wear caused by malicious damage by users; and generate a quantitative wear score based on normal wear and abnormal wear; The specific steps for distinguishing normal wear from abnormal wear in the wear analysis module 3 are as follows: H1. Extract the position distribution, shape regularity and depth difference data of each wear candidate area from the second three-dimensional wear distribution map, and count the area proportion and depth extreme value of the key areas of the forefoot, heel and arch of the sole; H2, compare the wear characteristics of the extracted wear candidate area with the standard wear data in the database module 4, and analyze whether the wear candidate area conforms to the distribution law of human gait mechanics. If the wear is concentrated in the common force area and the shape is naturally diffused, it is marked as normal wear; H3: Abnormal wear is determined when at least one of the following conditions is detected: a. The wear candidate area has regular geometric features and uniform depth variation, which is consistent with the tool damage characteristics; b. Isolated wear areas with sudden changes in depth appear in the non-stressed areas of the sole; c. The total area ratio of multiple wear areas on the same sole exceeds the normal threshold corresponding to the wearing time; d. The texture similarity between the wear candidate area and the vandalism cases in the malicious wear feature library exceeds 90%.

[0025] The steps for generating the quantitative wear score in the wear analysis module 3 are as follows: F1. According to the position, area percentage, depth difference data of the wear candidate areas marked in the second three-dimensional wear distribution map, as well as the normal wear and abnormal wear, the cumulative area percentage, maximum depth value, number of abnormal wear areas and distribution density parameters of each wear candidate area are extracted. Combined with the standard wear data of the same shoe type in the database module 4, dynamic weights are assigned to the wear parameters of different wear candidate areas, among which the area percentage weight of the forefoot and heel areas of the sole is higher than that of the arch area, and the weight of the abnormal wear area is 3 times that of the normal wear area; F2, calling the user's wearing history data in the database module 4, and adjusting the baseline threshold of the wear score according to the actual wearing time and usage scenario; F3. If there are wear candidate areas with abnormal wear or normal wear, an additional score deduction value is added based on the matching degree of its regular geometric features with the malicious wear feature library. The deduction value is proportional to the area ratio of the abnormal wear area, and inversely proportional to the area ratio of the normal wear area. F3. The wear parameters, baseline thresholds and deduction values ​​that have been assigned dynamic weights are superimposed to generate a quantitative wear score of 0-100 points, which is divided into four levels: 80 points or more for slight wear, 60-80 points for moderate wear, 30-60 points for severe wear and 30 points or less for severe damage.

[0026] The decision output module 5 is used to combine the quantitative wear score with the preset threshold value to output the quality assurance judgment conclusion and maintenance plan.

[0027] Among them, communication connections are established among the image acquisition module 1 , the image processing module 2 , the wear analysis module 3 , the database module 4 and the decision output module 5 .

[0028] The specific operation process of the decision output module 5 is as follows: C1. According to the level corresponding to the quantitative wear score, match the preset maintenance plan library and extract the maintenance cost corresponding to each level; C2. Based on the abnormal wear type, the following warranty judgment conclusions are obtained: If the abnormal wear is caused by non-uniform deformation due to sole quality problems, the conclusion "in compliance with warranty range" is output; if the abnormal wear is consistent with the regular geometric wear characteristics caused by malicious damage by the user, the conclusion "artificial damage" is output; if the quantitative wear score of the normal wear area is lower than the preset threshold but does not trigger an abnormal judgment, the conclusion "normal wear" is output; C3. Generate corresponding repair steps based on the repair cost and warranty judgment conclusion, integrate them into a visual report, mark the repair priority and cost details of each wear area, and output them to the preset terminal.

[0029] Example 3: After a pair of sports shoes of a certain brand is detected by the system, the wear degree analysis module 3 and the decision output module 5 execute the following process: Feature extraction: Extract the parameters of the forefoot, heel, and arch wear areas from the three-dimensional distribution data. Among them, the forefoot area accounts for 12%, the maximum depth difference is 1.5 mm, the heel accounts for 6%, the depth difference is 0.8 mm, the arch accounts for 2%, and the depth difference is 0.3 mm. And a regular circular wear area is found at the heel (diameter 8 mm, depth difference 1.2 mm); Abnormality determination: The matching degree between the regular circular wear area and the "repeated scraping with hard objects" case in the malicious wear library reaches 92%, and it is determined as man-made damage; An isolated wear point appears in the non-loading area of the arch (depth difference 0.5 mm), triggering the abnormal condition; Score calculation: Normal wear weight distribution: forefoot (12%×1.2), heel (6%×1.0), arch (2%×0.5); Deduction for abnormal areas: Regular wear area (area 0.5%×3 times the weight) + isolated point in the arch (0.1%×3 times), total deduction of 6 points; Considering the user's wearing duration (8 months), the benchmark threshold is adjusted to 70 points, and the final score = 78 (normal) - 6 (abnormal) = 72 points (moderate wear); Decision output: Warranty conclusion: The regular wear at the heel belongs to "man-made damage", and the isolated point in the arch belongs to "quality defect", partially meeting the warranty; Repair plan: Forefoot patch repair (￥80), heel replacement (￥150, at one's own expense), arch reinforcement (free under warranty), and generate a visual report and push it to the merchant terminal.

[0030] It should be noted that: The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: They can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; And these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A sole wear degree detection system based on image recognition, characterized in that Including: An image acquisition module (1), an image processing module (2), a wear degree analysis module (3), a database module (4), and a decision output module (5); The image acquisition module (1), which is used to obtain multi-angle high-definition images of the sole; The image processing module (2), which is used to remove impurities from the multi-angle high-definition images, generate a three-dimensional image model of the sole according to the multi-angle high-definition images through a preset three-dimensional reconstruction unit, and analyze the three-dimensional image model and the preset standard three-dimensional shape features of the sole through a preset processing unit to obtain the three-dimensional distribution data of the worn area of the sole; The database module (4), which is used to store the standard three-dimensional shape features of the soles of different shoe types, user wearing history data, a malicious wear feature library, and standard wear data; The wear degree analysis module (3), which performs the following determinations based on a pre-trained deep learning model: distinguishing normal wear and abnormal wear according to the database module (4) and the three-dimensional distribution data, where the abnormal wear includes non-uniform deformation caused by sole quality problems and regular geometric wear caused by user malicious damage; and generating a quantitative wear score according to the normal wear and the abnormal wear; The decision output module (5), which is used to combine the quantitative wear score and a preset threshold to output a quality assurance determination conclusion and a repair plan.

2. The sole wear degree detection system based on image recognition according to claim 1, characterized in that: Communication connections are established between the image acquisition module (1), the image processing module (2), the wear degree analysis module (3), the database module (4), and the decision output module (5); The image acquisition module (1) includes a rotatable stage, an annular LED light source array, and a high-resolution industrial camera.

3. The sole wear degree detection system based on image recognition according to claim 2, characterized in that: The specific steps for removing impurities from the multi-angle high-definition images in the image processing module (2) are as follows: S1. Calibrate the positions of the multi-angle high-definition images of the sole to eliminate image offsets caused by shooting angle differences. Through color recognition and contour extraction techniques, separate the sole body and the shooting background in the multi-angle high-definition images of the sole, delete the interference pixels of the rotatable stage, shadows, and non-sole areas, and retain multiple first complete sole images; S2. For the bright spots or specular effects generated by light reflection in the multiple first complete sole images, dynamically adjust the parameters of the annular LED light source array and the image brightness and darkness balance to eliminate the interference of the high-light area on the texture details. And through a preset repair unit, automatically identify the missing areas in some regions of the multiple first complete sole images caused by sole stains or shooting angle limitations, predict and fill the image content of the missing areas through the texture features of adjacent regions, and generate multiple second complete sole images; S3. Reduce the random noise in the multiple second complete sole images through a multi-level filtering method, smooth the granular noise caused by environmental interference, and strengthen the contour features of the sole pattern edges and worn areas, highlighting the transition difference between the worn boundary and the normal area, and completing the impurity removal process of the multi-angle high-definition images of the sole.

4. The sole wear degree detection system based on image recognition according to claim 3, wherein: The three-dimensional distribution data of the sole wear area in the image processing module (2) is obtained through the following steps: Q1. The three-dimensional reconstruction unit uses multiple processed second complete sole images to extract the spatial position information of the sole from different perspectives, calculates and generates a set of three-dimensional spatial data points on the sole surface through geometric relationships, and connects the spatial relationships between adjacent three-dimensional spatial data points according to the distribution law of the three-dimensional spatial data point set to form a continuous three-dimensional image model of the sole; Q2. The processing unit performs mirror processing on the three-dimensional image model and the standard three-dimensional shape features of the sole, and preset key point alignment methods are used to eliminate the spatial deviation caused by the shooting angle. A preset reference datum plane is marked in the middle of the three-dimensional image model and the standard three-dimensional shape features of the sole. The vertical distances between the three-dimensional image model and the reference datum plane, and between the standard three-dimensional shape features of the sole and the reference datum plane are calculated point by point to generate a first set of depth change values and a second set of depth change values; Q3. The first set of depth change values is compared with the second set of depth change values, and the values in the first set of depth change values that exceed or are less than the values in the second set of depth change values are analyzed to generate a first three-dimensional wear distribution map of depth differences. A second three-dimensional wear distribution map including positions and area ratios is generated based on the first three-dimensional wear distribution map, and the wear parameters of the key areas of the front sole, heel, and arch of the sole are marked in the first three-dimensional wear distribution map and the second three-dimensional wear distribution map to obtain the three-dimensional distribution data of the sole wear area.

5. The sole wear degree detection system based on image recognition according to claim 4, wherein: The specific steps for generating the second three-dimensional wear distribution map including positions and area ratios based on the first three-dimensional wear distribution map in step Q3 are as follows: Q31. Based on the depth difference data of the first three-dimensional wear distribution map, areas with depth changes exceeding a preset threshold are delimited on the surface of the three-dimensional image model, marked as wear candidate areas, and the central coordinates and boundary ranges of each wear candidate area are recorded; Q32. For each wear candidate area, the surface area of the sole covered by it is calculated according to its boundary range, and the ratio is converted with the overall surface area of the sole to obtain the area ratio value of each wear candidate area; Q33. According to the combined results of the depth difference data of the first three-dimensional wear distribution map and the area ratio values, the wear candidate areas are divided into three wear levels: mild, moderate, and severe; Q34. In the surface coordinate system of the three-dimensional image model, a circular coverage area is generated with the central coordinates of the wear candidate area as the anchor point, and the radius size is proportional to the area ratio value, and the color depth of the circular coverage area is positively correlated with the wear level; Q35. The area ratio values of each wear candidate area are marked at the center of the corresponding circular coverage area, and adjacent wear candidate areas are connected by connection lines to form a second three-dimensional wear distribution map showing the wear distribution density and severity.

6. The sole wear degree detection system based on image recognition according to claim 5, characterized in that: The specific steps for distinguishing normal wear from abnormal wear in the wear analysis module (3) are: H1, extracting the position distribution, shape regularity and depth difference data of each wear candidate area from the second three-dimensional wear distribution map, and counting the area proportion and depth extreme value of the key areas of the forefoot, heel and arch of the sole; H2, comparing the extracted wear characteristics of the candidate wear area with the standard wear data in the database module (4), analyzing whether the candidate wear area conforms to the distribution law of human gait mechanics; if the wear is concentrated in the common force-bearing area and the shape is naturally diffused, it is marked as normal wear; H3: Abnormal wear is determined when at least one of the following conditions is detected: a. The wear candidate area has regular geometric features and uniform depth variation, which is consistent with the tool damage characteristics; b. Isolated wear areas with sudden changes in depth appear in the non-stressed areas of the sole; c. The total area ratio of multiple wear areas on the same sole exceeds the normal threshold corresponding to the wearing time; d. The texture similarity between the wear candidate area and the human damage cases in the malicious wear feature library exceeds 90%.

7. The sole wear degree detection system based on image recognition according to claim 6, characterized in that: The steps for generating the quantitative wear score in the wear analysis module (3) are as follows: F1, based on the location, area percentage, depth difference data of the wear candidate areas marked in the second three-dimensional wear distribution map, as well as the normal wear and the abnormal wear, extract the cumulative area percentage, maximum depth value, number of abnormal wear areas and distribution density parameters of each wear candidate area, and combine the standard wear data of the same shoe type in the database module (4) to assign dynamic weights to the wear parameters of different wear candidate areas, wherein the area percentage weights of the forefoot and heel areas of the sole are higher than those of the arch area, and the weight of the abnormal wear area is three times that of the normal wear area; F2, calling the user's wearing history data in the database module (4), and adjusting the baseline threshold of the wear score according to the actual wearing time and usage scenario; F3. If there is a candidate wear area with abnormal wear or normal wear, an additional score deduction value is added according to the matching degree of its regular geometric features with the malicious wear feature library. The deduction value is proportional to the area ratio of the abnormal wear area, and the deduction value is inversely proportional to the area ratio of the normal wear area; F3. The wear parameters assigned with dynamic weights, the baseline threshold and the deduction value are superimposed to generate a quantitative wear score of 0-100 points, which is divided into four levels: 80 points or more for slight wear, 60-80 points for moderate wear, 30-60 points for severe wear and 30 points or less for serious damage.

8. The sole wear degree detection system based on image recognition according to claim 7, wherein: The specific operation process of the decision output module (5) is as follows: C1. According to the level corresponding to the quantitative wear score, a preset maintenance solution library is matched to extract the maintenance cost corresponding to each level; C2. Based on the abnormal wear type, the following warranty judgment conclusion is obtained: If the abnormal wear is caused by non-uniform deformation resulting from sole quality problems, output the conclusion of "within the warranty scope"; if the abnormal wear conforms to the regular geometric wear characteristics caused by malicious damage by the user, output the conclusion of "man-made damage"; if the quantified wear score in the normal wear area is lower than the preset threshold but does not trigger an abnormal determination, output the conclusion of "normal wear and tear". C3. Generate corresponding repair steps according to the repair cost and the warranty determination conclusion, integrate them into a visual report, mark the repair priority and cost details of each wear area, and output them to a preset terminal.

Images