Intelligent detection system and method for appearance defects of sports shoes based on small sample learning

Abstract

The present application relates to the technical field of visual detection, and particularly relates to a sneaker appearance flaw intelligent detection system and method based on small sample learning, which comprises a flaw intelligent detection center, a flaw analysis unit, a checking and obtaining unit, a difficulty evaluation unit and a visual feedback unit; the present application realizes the flaw type identification and the boundary box coordinate output of the sneaker through the combined mode of the model and the 2D image, and simultaneously carries out the precision analysis of the boundary box coordinates, so as to provide data support for the subsequent, obtains the flaw area based on the boundary box coordinate analysis, simultaneously carries out the 3D and spectral image analysis on the flaw area, checks the obtained flaw type with the flaw type analyzed by the 2D image, so as to ensure the consistency of the flaw result, reduce the error rate of the flaw detection result, and simultaneously carries out the repair feasibility discrimination processing, which is helpful for the repair difficulty classification of the sneaker with the flaw, that is, the objective classification is realized by combining the multi-index weighted fusion.

Description

Technical Field

[0001] This invention relates to the field of visual inspection technology, and in particular to an intelligent system and method for detecting appearance defects in athletic shoes based on few-sample learning. Background Technology

[0002] Appearance quality inspection of athletic shoes is a key process in the production process, which directly affects product quality and brand image. In the existing technology, deep learning-based detection methods require a large number of labeled samples to train the model. However, athletic shoes have a fast update cycle (thousands of styles every year), a variety of materials (leather, mesh, TPU, etc.), and a scarcity of samples of new defects (such as interlayer bubbles in 3D printed shoe uppers), resulting in insufficient generalization ability of the model.

[0003] Few-shot learning technology can quickly master new tasks from a small number of samples by learning meta-knowledge of "how to learn", providing a new approach to solving the problem of sample scarcity. However, traditional systems rely solely on 2D image detection, which makes it difficult to accurately identify defects and their locations, resulting in low detection accuracy. At the same time, it is difficult to verify the accuracy of location information, leading to excessive deviation in the subsequent defect identification accuracy analysis results, which is not conducive to the defect repair and classification management of sports shoes.

[0004] To address the aforementioned technical shortcomings, a solution is proposed. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent detection system and method for sports shoe appearance defects based on few-sample learning, in order to solve the aforementioned technical deficiencies. This invention achieves defect type identification and bounding box coordinate output for sports shoes by combining models and 2D images, while simultaneously performing accuracy analysis of the bounding box coordinates to provide data support for subsequent processing. Furthermore, 3D and spectral image analysis is performed on the defect area, and the obtained defect types are compared with those obtained from 2D image analysis to ensure consistency of defect results and reduce the error rate of defect detection results. Additionally, repair feasibility discrimination processing is included, which helps to classify the repair difficulty of sports shoes with defects, i.e., achieving objective classification through multi-indicator weighted fusion.

[0006] The objective of this invention can be achieved through the following technical solution: a sports shoe appearance defect intelligent detection system based on few-sample learning, including a defect intelligent detection center, a defect analysis unit, a verification and acquisition unit, a difficulty assessment unit, and a visual feedback unit;

[0007] The intelligent defect detection center is used to retrieve and store the surface feature image set of sports shoes;

[0008] The defect analysis unit is used to identify and locate defects in the surface feature image set to obtain the first defect type and bounding box coordinates.

[0009] A pixel-level feature comparison and analysis process based on bounding box coordinate accuracy verification yields standard or trimmed signals.

[0010] When a standard signal is generated, the verification and acquisition unit is used to perform coordinate transformation and defect identification analysis on the bounding box coordinates, and performs consistency comparison analysis between the obtained second defect type and the first defect type to obtain a self-test signal or output signal.

[0011] The difficulty assessment unit is used to quantitatively assess and analyze the repair difficulty of the surface images of the defective areas in the collected sports shoes, and to determine whether the obtained repair difficulty coefficient exceeds the preset repair difficulty coefficient threshold, so as to obtain a difficult repair signal or a repairable signal.

[0012] Preferably, the surface feature image set represents images of the front, side, and sole of the athletic shoe obtained by using an industrial camera and light source assembly.

[0013] Preferably, the defect identification and location analysis process is as follows:

[0014] S1: Preprocess the acquired surface feature image set, including noise reduction and size normalization;

[0015] S2: Extract the surface feature image set of the sports shoes, manually annotate the defects in the surface feature image set, and the annotation content includes the defect type and location information to obtain the sample training dataset containing manual annotations;

[0016] S3: Preprocess the images in the sample training dataset to obtain preprocessed training samples. Divide the training samples into support set and query set. Iteratively train based on support set and query set to finally build a defect recognition model.

[0017] S4: Input the preprocessed surface feature image set into the defect recognition model to obtain the first defect type and bounding box coordinates output by the defect recognition model.

[0018] The preferred pixel-level feature comparison and analysis process is as follows:

[0019] The appearance feature images of various historical defects and the bounding boxes and feature labels marked in the appearance feature images are obtained. Based on the bounding boxes and feature labels marked in the appearance feature images, a defect feature library is constructed after preprocessing. The defect feature library includes morphological features, color and texture features.

[0020] The pixel features within the bounding box are obtained based on the bounding box coordinates. The pixel features include texture and color. The extracted pixel features are compared with the defect feature library to obtain the similarity between the pixel features within the bounding box and the defect feature library. The similarity is then judged to see if it exceeds a preset similarity threshold. If it does, a standard signal is generated; otherwise, a trimming signal is generated.

[0021] Preferably, the coordinate transformation and defect identification analysis process is as follows:

[0022] The obtained bounding box coordinates are converted into 3D point cloud coordinates using existing technology, and the area enclosed by the 3D point cloud coordinates is set as the defect area.

[0023] Point cloud image information of the defective area in the sports shoe was obtained by using a 3D structured light scanner, and surface spectral image data of the defective area in the sports shoe was obtained by using a hyperspectral camera.

[0024] Preprocessing is performed on point cloud image information and surface spectral image data, and feature extraction is performed on the preprocessed point cloud image information and surface spectral image data respectively;

[0025] The feature stitching method in the existing technology is used to fuse the features of the aligned point cloud image information and the features of the surface spectral image data to obtain a multimodal feature vector.

[0026] Preferably, the multimodal feature vectors are input into a pre-set meta-knowledge graph model to obtain the output geometric similarity and spectral similarity.

[0027] The sum of the geometric similarity and spectral similarity multiplied by their respective preset weight coefficients is set as the comprehensive similarity. The comprehensive similarity is then judged to see if it exceeds the preset comprehensive similarity threshold. If it does, the second defect type is output; otherwise, an artificial signal is generated.

[0028] Preferably, a consistency comparison analysis is performed on the first defect type and the second defect type. If the first defect type and the second defect type are consistent, an output signal is generated; if the first defect type and the second defect type are inconsistent, a self-test signal is generated.

[0029] Preferably, the quantitative assessment and analysis process for the repair difficulty is as follows:

[0030] The surface image of the defective area in the sports shoe is obtained, and the surface image is processed into grayscale. The total number of defective pixels within the defect boundary box in the grayscale processed surface image is obtained.

[0031] The difference between the average gray value in the defective area and the preset average gray value threshold is obtained, and the difference between the average gray value in the defective area and the preset average gray value threshold is set as the pixel gray value difference;

[0032] The main skeleton of the defect within the defect area is obtained by using a skeleton extraction algorithm (such as Zernike moments), the pixel distance between the two ends of the skeleton is calculated, and the pixel distance between the two ends of the skeleton is set as the defect length.

[0033] The total number of defective pixels, pixel grayscale difference, and defect length are multiplied by the corresponding set weight coefficients, and the sum of the multiplication is set as the repair difficulty coefficient. The repair difficulty coefficient is then judged to see if it exceeds the preset repair difficulty coefficient threshold. If it does, a difficult repair signal is generated; otherwise, a repairable signal is generated.

[0034] The beneficial effects of this invention are as follows:

[0035] (1) This invention adopts small sample learning technology, which only requires a small number of labeled samples to train the model, greatly reducing the dependence on a large amount of labeled data and reducing the cost of manual labeling. At the same time, it realizes the identification of the defect type of sports shoes and the output of bounding box coordinates by combining the model and 2D images. Furthermore, it performs pixel-level feature comparison analysis on the accuracy of the bounding box coordinates to determine whether the accuracy of the bounding box coordinates is qualified, so as to provide data support for the subsequent analysis and improve the accuracy of the subsequent analysis results.

[0036] (2) The present invention obtains the defect area based on bounding box coordinate analysis, and performs 3D and spectral image analysis on the defect area. The defect type obtained is checked with the defect type of 2D image analysis to ensure the consistency of defect results and reduce the error rate of defect detection results.

[0037] (3) The present invention conducts an in-depth assessment of the repair difficulty from the perspective of pixel-level features of the defective area, and is accompanied by repair feasibility discrimination processing, which helps to classify the repair difficulty of sports shoes with defects, that is, to achieve objective classification by combining multiple indicators weighted fusion. Attached Figure Description

[0038] The invention will now be further described with reference to the accompanying drawings;

[0039] Figure 1 This is a flowchart of the system of the present invention;

[0040] Figure 2 This is a reference diagram for the analysis of the method of this invention;

[0041] Figure 3 This is a partial analysis diagram of Embodiment 1 of the present invention. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments;

[0044] Example 1:

[0045] Please see Figures 1 to 3 As shown, the present invention is an intelligent detection system for sports shoe appearance defects based on few-sample learning, including a defect intelligent detection center, a defect analysis unit, a verification and acquisition unit, a difficulty assessment unit, and a visual feedback unit. The defect intelligent detection center is connected to the defect analysis unit and the visual feedback unit in a one-way communication. The defect analysis unit is connected to the verification and acquisition unit in a one-way communication. The verification and acquisition unit is connected to the difficulty assessment unit in a one-way communication. The difficulty assessment unit is connected to the defect intelligent detection center in a one-way communication.

[0046] The intelligent defect detection center is used to retrieve and store the surface feature image set of sports shoes;

[0047] The defect analysis unit is used to identify and locate defects in the surface feature image set. The specific defect identification and location analysis process is as follows:

[0048] The surface feature image set represents images of the athletic shoe from multiple angles, such as the front, side, and sole, obtained by using industrial cameras, light source components, and other equipment.

[0049] S1: Preprocess the acquired surface feature image set, including noise reduction and size normalization;

[0050] For example: denoising uses Gaussian filtering to remove noise from the image; size normalization adjusts the image to a preset size (such as 224×224 pixels) to facilitate model processing.

[0051] S2: Extract the surface feature image set of the sports shoes, manually annotate the defects in the surface feature image set, and the annotation content includes the defect type (scratches, stains, glue separation, color difference, etc.) and location information (bounding box coordinates) to obtain a sample training dataset containing manual annotations.

[0052] S3: Preprocess the images in the sample training dataset to obtain preprocessed training samples. Divide the training samples into support set and query set. Iteratively train based on support set and query set to finally build a defect recognition model.

[0053] S4: Input the preprocessed surface feature image set into the defect recognition model to obtain the first defect type and bounding box coordinates output by the defect recognition model;

[0054] S5: Pixel-level feature comparison and analysis process based on bounding box coordinate accuracy verification: Obtain the appearance feature images of various historical defects and the bounding boxes, feature labels and other information marked in the appearance feature images. After preprocessing based on the bounding boxes, feature labels and other information marked in the appearance feature images, construct a defect feature library. The defect feature library includes morphological features, color and texture features, etc.

[0055] The pixel features within the bounding box, including texture and color, are obtained based on the bounding box coordinates. These pixel features are then compared with a defect feature library to obtain the similarity between the pixel features within the bounding box and the defect feature library. The similarity is then judged to see if it exceeds a preset similarity threshold. If it does, a standard signal is generated; otherwise, a correction signal is generated. The visual feedback unit responds to the standard signal or the correction signal and immediately displays the preset warning text corresponding to the standard signal or the correction signal to revise the obtained bounding box coordinates and improve their accuracy.

[0056] Example 2:

[0057] When generating the standard signal, the verification and acquisition unit is used to perform coordinate transformation and defect identification analysis on the bounding box coordinates. The specific coordinate transformation and defect identification analysis process is as follows:

[0058] The bounding box coordinates (2D) of the image are based on the pixel coordinate system (the origin is at the top left corner of the image, and the unit is pixels), while the 3D point cloud coordinates are based on the world coordinate system (or camera coordinate system, and the unit is millimeters / meters). The conversion between the two depends on the camera intrinsic parameters (describing the camera's optical characteristics) and extrinsic parameters (describing the pose relationship between the camera and the world coordinate system), and the mapping is achieved through the perspective projection model.

[0059] The obtained bounding box coordinates are converted into 3D point cloud coordinates using existing technology, and the area enclosed by the 3D point cloud coordinates is set as the defect area.

[0060] Point cloud image information of the defective area in the sports shoe was obtained by using a 3D structured light scanner, and surface spectral image data of the defective area in the sports shoe was obtained by using a hyperspectral camera.

[0061] Preprocessing is performed on point cloud image information and surface spectral image data, including noise reduction and smoothing. Feature extraction is then performed on the preprocessed point cloud image information and surface spectral image data to obtain the features of the preprocessed point cloud image information, such as curvature change rate and normal vector deviation, and the features of the preprocessed surface spectral image data, such as spectral angle and reflectance deviation.

[0062] By employing the existing SIFT (Scale Invariant Feature Transform) feature point matching algorithm and ICP (Iterative Closest Point) algorithm, coordinate alignment between 3D point cloud images and hyperspectral images can be achieved. In this way, the positions of the two can be quickly aligned.

[0063] Through precise feature alignment, the effective fusion of 3D geometric features and hyperspectral features was achieved, giving full play to the complementary advantages of multimodal data and providing more comprehensive feature support for subsequent defect detection.

[0064] The feature stitching method in the existing technology is used to fuse the features of the aligned point cloud image information and the features of the surface spectral image data to obtain a multimodal feature vector;

[0065] The extracted geometric features, such as the rate of curvature change and the deviation of the normal vector, are concatenated with spectral features, such as the spectral angle and the reflectance deviation, to form a feature vector with higher dimensions and richer information, providing input for the small sample fusion detection model.

[0066] The multimodal feature vectors are input into a pre-set meta-knowledge graph model to obtain the output geometric similarity (the spatial difference between the geometric features of athletic shoes and standard athletic shoes is measured by Euclidean distance) and spectral similarity (the angular difference between the spectral features of athletic shoes and standard athletic shoes is measured by cosine similarity).

[0067] A pre-defined meta-knowledge graph model stores the "geometric-spectral" correlation patterns of defects in the form of triples, for example:

[0068] Scratches: Geometric features are "linear depressions, depth 0.05-0.3 mm, length > 1 mm", spectral features are "reflectivity decrease of 15%-30% in the 450 nm band"; Bubbles: Geometric features are "circular protrusions, height 0.05-0.2 mm, diameter 0.5-3 mm", spectral features are "deterioration in reflectivity uniformity across the entire band, ΔR > 10%", etc.

[0069] The sum of the geometric similarity and spectral similarity multiplied by the corresponding preset weight coefficients is set as the comprehensive similarity. The comprehensive similarity is then judged to see if it exceeds the preset comprehensive similarity threshold. If it does, the second defect type is output; otherwise, an artificial signal is generated.

[0070] The first defect type and the second defect type are compared and analyzed for consistency. If the first defect type and the second defect type are consistent, an output signal is generated. If the first defect type and the second defect type are inconsistent, a self-test signal is generated. The visual feedback unit is used to respond to the manual signal, the self-test signal or the output signal, and immediately display the preset warning text corresponding to the manual signal, the self-test signal or the output signal. This allows for an intuitive understanding of the defect type and bounding box coordinates of the sports shoe based on the feedback information, as well as an understanding of the error of the defect detection results. This enables timely feedback and warning management to ensure the accuracy and reliability of defect detection in sports shoes.

[0071] The difficulty assessment unit is used to quantitatively assess and analyze the repair difficulty of surface images of defective areas in the acquired sports shoes. The specific process of quantitative assessment and analysis of repair difficulty is as follows:

[0072] The surface image of the defective area in the sports shoe is obtained, and the surface image is processed into grayscale. The total number of defective pixels within the defect boundary box in the grayscale processed surface image is obtained.

[0073] The difference between the average gray value in the defective area and the preset average gray value threshold is obtained, and the difference between the average gray value in the defective area and the preset average gray value threshold is set as the pixel gray value difference;

[0074] The main skeleton of the defect within the defect area is obtained by using a skeleton extraction algorithm (such as Zernike moments), the pixel distance between the two ends of the skeleton is calculated, and the pixel distance between the two ends of the skeleton is set as the defect length.

[0075] The total number of defective pixels, pixel grayscale difference, and defect length are multiplied by their respective weighting coefficients. The sum of these multiplications is set as the repair difficulty coefficient. The repair difficulty coefficient is then judged to see if it exceeds the preset repair difficulty coefficient threshold. If it does, a difficult-to-repair signal is generated; otherwise, a repairable signal is generated. The visual feedback unit responds to either the difficult-to-repair or repairable signal and immediately displays the preset warning operation corresponding to the signal. This helps to classify the repair difficulty of defective athletic shoes, achieving objective classification through the weighted fusion of multiple indicators.

[0076] Example 3:

[0077] A method for intelligent detection of appearance defects in athletic shoes based on few-sample learning includes the following steps:

[0078] Step 1: Retrieving and storing the surface feature image set of the athletic shoe;

[0079] Step 2: Defect identification and localization analysis based on surface feature image set, that is, performing defect identification and localization analysis on surface feature image set to obtain the first defect type and bounding box coordinates;

[0080] Step 3: Pixel-level feature comparison and analysis process based on bounding box coordinate accuracy verification, that is, comparing and analyzing the pixel features within the obtained bounding box to obtain standard signals or trimmed signals;

[0081] Step 4: Based on the information progression, the bounding box coordinate transformation and defect identification analysis process is carried out, that is, whether the obtained comprehensive similarity exceeds the preset comprehensive similarity threshold is judged and processed to obtain artificial signal, self-inspection signal or output signal;

[0082] Step 5: Perform a quantitative assessment and analysis of the repair difficulty of the surface image of the defective area in the sports shoe, that is, determine whether the obtained repair difficulty coefficient exceeds the preset repair difficulty coefficient threshold, and obtain a difficult repair signal or a repairable signal.

[0083] In summary, this invention employs few-sample learning technology, requiring only a small number of labeled samples to train the model, significantly reducing reliance on large amounts of labeled data and lowering manual annotation costs. It combines the model with 2D images to identify defect types and output bounding box coordinates for athletic shoes. Furthermore, it performs pixel-level feature comparison analysis to determine the accuracy of the bounding box coordinates, providing data support for subsequent analysis and improving accuracy. Based on the bounding box coordinate analysis, defect areas are identified, and 3D and spectral image analysis is performed on these areas. The resulting defect types are compared with those from the 2D image analysis to ensure consistency and reduce the error rate of defect detection. The invention also provides in-depth pixel-level feature assessment of the defect area, along with repair feasibility discrimination, facilitating the classification of repair difficulty for defective athletic shoes. This multi-indicator weighted fusion approach achieves objective classification.

[0084] The threshold is set for comparative analysis of results to determine whether they are good or bad. The value of the threshold is determined by a combination of large-scale model analysis of sample data and human experience. It can also be adjusted appropriately based on seasonal or common-sense influencing factors.

[0085] The size of the coefficient is a specific value obtained by quantifying each parameter to facilitate subsequent comparison. The size of the coefficient depends on the amount of sample data and the corresponding operating coefficient initially set by those skilled in the art for each set of sample data; as long as it does not affect the proportional relationship between the parameter and the quantified value.

[0086] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A smart detection system for appearance defects in athletic shoes based on few-sample learning, characterized in that, It includes a defect intelligent detection center, a defect analysis unit, a verification and acquisition unit, a difficulty assessment unit, and a visual feedback unit; The intelligent defect detection center is used to retrieve and store the surface feature image set of sports shoes; The defect analysis unit is used to identify and locate defects in the surface feature image set to obtain the first defect type and bounding box coordinates. A pixel-level feature comparison and analysis process based on bounding box coordinate accuracy verification yields standard or trimmed signals. When a standard signal is generated, the verification and acquisition unit performs coordinate transformation and defect identification analysis on the bounding box coordinates, compares the obtained second defect type with the first defect type for consistency analysis, and obtains a self-test signal or an output signal; the coordinate transformation and defect identification analysis process is as follows: The obtained bounding box coordinates are converted into 3D point cloud coordinates using existing technology, and the area enclosed by the 3D point cloud coordinates is set as the defect area. Point cloud image information of the defective area in the sports shoe was obtained by using a 3D structured light scanner, and surface spectral image data of the defective area in the sports shoe was obtained by using a hyperspectral camera. Preprocessing is performed on point cloud image information and surface spectral image data, and feature extraction is performed on the preprocessed point cloud image information and surface spectral image data respectively; The feature stitching method in the existing technology is used to fuse the features of the aligned point cloud image information and the features of the surface spectral image data to obtain a multimodal feature vector; the multimodal feature vector is input into a pre-set meta-knowledge graph model to obtain the output geometric similarity and spectral similarity; The sum of the geometric similarity and spectral similarity multiplied by the corresponding preset weight coefficients is set as the comprehensive similarity. The comprehensive similarity is then judged to see if it exceeds the preset comprehensive similarity threshold. If it does, the second defect type is output; otherwise, an artificial signal is generated. The difficulty assessment unit is used to quantitatively assess and analyze the repair difficulty of the surface images of the defective areas in the collected sports shoes, and to determine whether the obtained repair difficulty coefficient exceeds the preset repair difficulty coefficient threshold, so as to obtain a difficult repair signal or a repairable signal.

2. The intelligent detection system for sports shoe appearance defects based on few-sample learning according to claim 1, characterized in that, The surface feature image set represents images of the front, side, and sole of the athletic shoe acquired using an industrial camera and light source assembly.

3. The intelligent detection system for sports shoe appearance defects based on few-sample learning according to claim 1, characterized in that, The defect identification and location analysis process is as follows: S1: Preprocess the acquired surface feature image set, including noise reduction and size normalization; S2: Extract the surface feature image set of the sports shoes, manually annotate the defects in the surface feature image set, and the annotation content includes the defect type and location information to obtain the sample training dataset containing manual annotations; S3: Preprocess the images in the sample training dataset to obtain preprocessed training samples. Divide the training samples into support set and query set. Iteratively train based on support set and query set to finally build a defect recognition model. S4: Input the preprocessed surface feature image set into the defect recognition model to obtain the first defect type and bounding box coordinates output by the defect recognition model.

4. The intelligent detection system for sports shoe appearance defects based on few-sample learning according to claim 1, characterized in that, The pixel-level feature comparison and analysis process is as follows: The appearance feature images of various historical defects and the bounding boxes and feature labels marked in the appearance feature images are obtained. Based on the bounding boxes and feature labels marked in the appearance feature images, a defect feature library is constructed after preprocessing. The defect feature library includes morphological features, color and texture features. The pixel features within the bounding box are obtained based on the bounding box coordinates. The pixel features include texture and color. The extracted pixel features are compared with the defect feature library to obtain the similarity between the pixel features within the bounding box and the defect feature library. The similarity is then judged to see if it exceeds a preset similarity threshold. If it does, a standard signal is generated; otherwise, a trimming signal is generated.

5. The intelligent detection system for sports shoe appearance defects based on few-sample learning according to claim 1, characterized in that, A consistency comparison analysis is performed on the first defect type and the second defect type. If the first defect type and the second defect type are consistent, an output signal is generated. If the first defect type and the second defect type are inconsistent, a self-test signal is generated.

6. The intelligent detection system for sports shoe appearance defects based on few-sample learning according to claim 1, characterized in that, The quantitative assessment and analysis process for the repair difficulty is as follows: The surface image of the defective area in the sports shoe is obtained, and the surface image is processed into grayscale. The total number of defective pixels within the defect boundary box in the grayscale processed surface image is obtained. The difference between the average gray value in the defective area and the preset average gray value threshold is obtained, and the difference between the average gray value in the defective area and the preset average gray value threshold is set as the pixel gray value difference; The main skeleton of the defect within the defect area is obtained by using the Zernike moment skeleton extraction algorithm. The pixel distance between the two ends of the skeleton is calculated and set as the defect length. The total number of defective pixels, pixel grayscale difference, and defect length are multiplied by the corresponding set weight coefficients, and the sum of the multiplication is set as the repair difficulty coefficient. The repair difficulty coefficient is then judged to see if it exceeds the preset repair difficulty coefficient threshold. If it does, a difficult repair signal is generated; otherwise, a repairable signal is generated.

7. A few-shot learning-based intelligent detection system for sports shoe appearance defects, wherein the method is applied to any one of claims 1-6, characterized in that, Includes the following steps: Step 1: Retrieving and storing the surface feature image set of the athletic shoe; Step 2: Defect identification and localization analysis based on surface feature image set Step 3: Pixel-level feature comparison and analysis process based on bounding box coordinate accuracy verification; Step 4: Coordinate transformation and defect identification analysis of bounding box coordinates based on information progression; Step 5: Quantitatively assess and analyze the difficulty of repairing the surface images of the defective areas in the sneakers.

Images