Face cover cutting piece quality detection method and system

Abstract

The invention relates to the technical field of cover detection, in particular to a cover cutting piece quality detection method and system, and the method comprises the steps: carrying out the image collection of a cutting piece, and carrying out the preprocessing; according to the shape features and edge quality requirements of the cover, identifying the edge of the cutting piece based on a self-adaptive edge detection algorithm; constructing an edge defect identification model, and performing defect identification and classification on the detected edge contour to obtain an edge defect detection result; for defect types which cannot be determined by edge detection and defect identification, uploading the cut-part image data and the preliminary analysis result to a cloud to obtain a cloud analysis result; and generating a final quality detection result according to an edge defect detection result and a cloud analysis result. According to the invention, the problems of low manual detection efficiency, poor accuracy and incomplete defect identification in the traditional cut-part quality detection are effectively solved, the edges and defects of the cut-parts are accurately identified, the detection precision and consistency are improved, and the manual error is reduced.

Description

Technical Field

[0001] The invention relates to the technical field of face cover detection, and in particular to a face cover cut piece quality detection method and system. Background Art

[0002] In the production process of car seat covers, the quality of the cut pieces directly affects the overall appearance and service life of the final product. At present, the quality inspection of most cover cut pieces still relies on manual visual inspection or simple automated equipment. This method has obvious limitations in inspection efficiency and accuracy. Especially when inspecting defects on the edges of cut pieces (such as burrs, unevenness, damage, etc.), manual inspection is often not accurate or consistent enough, resulting in uneven quality of finished products and affecting market competitiveness.

[0003] The information disclosed in this background technology section is only intended to deepen the understanding of the overall background technology of the present disclosure, and should not be regarded as acknowledging or suggesting in any form that the information constitutes the prior art known to those skilled in the art. Summary of the invention

[0004] The present invention provides a method and system for detecting the quality of a cover piece, thereby effectively solving the problems pointed out in the background technology.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is: A method for detecting the quality of a cover piece, comprising: Capture images of the cut pieces in the inspection area, obtain high-definition images of the cut pieces under standard lighting conditions, and perform preprocessing; According to the shape characteristics and edge quality requirements of the face cover, based on the adaptive edge detection algorithm, the edge detection parameters suitable for the material and cutting process of the cut piece are selected, and the gradient threshold and operator scale are dynamically adjusted to identify the edge of the cut piece; Construct an edge defect recognition model, and based on the edge contour data of the cut piece, perform defect recognition and classification on the detected edge contour to obtain the edge defect detection result; For defect types that cannot be determined by edge detection and defect identification, upload the cut piece image data and preliminary analysis results to the cloud to obtain cloud analysis results; Based on the edge defect detection results and cloud analysis results, the comprehensive quality score of the cut pieces is calculated to generate the final quality inspection results.

[0006] Furthermore, the preprocessing includes: Grayscale processing is performed on the cut pieces of different materials to obtain grayscale images; Apply a lighting correction algorithm based on the reflective properties of the cut material to obtain an image with uniform lighting; A filtering method is used to remove noise from the image to obtain a denoised image; Image enhancement technology is used to improve image contrast and obtain an enhanced image.

[0007] Furthermore, the illumination correction algorithm includes: Performing a histogram analysis on each of the grayscale images to determine the distribution of the brightness of the grayscale images; Based on the local adaptive histogram equalization method, the brightness of different areas of the image is adjusted; Construct a reflectivity correction model and apply multispectral imaging technology to compensate for reflective areas, and use shadow monitoring and interpolation methods to correct shadow parts in the image; Adjust and normalize the global brightness, and set the brightness distribution of the image within a standard range through linear stretching and standardization algorithms; The lighting correction effect is evaluated and the parameters are corrected based on the evaluation results.

[0008] Furthermore, constructing the reflectivity correction model includes: Obtain the optical properties of each material of the cover piece under different light sources and build a reflection property database; Determine the key parameters of the model and construct a Phong-based illumination model to describe the reflection behavior of the material surface; Based on the illumination model and the reflection characteristic database, a reflectivity correction formula is constructed to calculate the brightness compensation value of the reflective area; The illumination model is tested using actual cut piece image data, the error of the correction effect is calculated, and the model parameters are adjusted to obtain the final reflectivity correction model.

[0009] Furthermore, according to the shape characteristics and edge quality requirements of the face cover, based on the adaptive edge detection algorithm, edge detection parameters suitable for the material and cutting process of the cut piece are selected, and the gradient threshold and operator scale are dynamically adjusted to identify the edge of the cut piece, including: Determine the shape characteristics and edge quality requirements of the cut piece, and select edge detection parameters suitable for the cut piece material and cutting process, including the initial gradient threshold and operator scale; Applying an adaptive edge detection algorithm to the collected image data of the cut piece, dynamically adjusting the gradient threshold according to the local contrast and noise level of the image; Adaptively adjust the scale of the edge detection operator according to the thickness and curvature changes of the edge of the cut piece; Through the multi-scale edge detection algorithm, the edge detection results at different scales are integrated to generate a comprehensive edge recognition image; The generated edge recognition image is subjected to morphological processing to obtain the final edge recognition result.

[0010] Furthermore, through the multi-scale edge detection algorithm, the edge detection results at different scales are integrated to generate a comprehensive edge recognition image, including: Perform multi-scale decomposition on the cut image to generate multiple image versions of different scales. Use image decomposition technology to downsample the original image step by step to form image pyramids of different resolutions. Applying edge detection algorithms to the images of each scale respectively to obtain edge detection results at different scales; The edge detection results at different scales are integrated based on the fusion strategy to generate a comprehensive edge recognition image.

[0011] Furthermore, for defect types that cannot be determined by edge detection and defect identification, the cut piece image data and preliminary analysis results are uploaded to the cloud to obtain cloud analysis results, including: Analyze and classify the unidentified defect types based on the cloud-based deep learning model; Based on the defect classification data, determine the exact location and boundary of the defect, and generate the corresponding coordinate data and area size information; Calculate the severity of the defect based on the basic characteristic parameters of the defect and grade and mark the defect; According to the defect classification data and grading marks, corresponding processing instructions and defect identification parameters are obtained, and cloud analysis results are output.

[0012] Furthermore, based on the deep learning model in the cloud, the types of defects that cannot be determined are analyzed and classified, including: If the defect type cannot be determined, the relevant image data of the cut piece and the preliminary analysis results are uploaded to the cloud, including the high-definition image of the cut piece, edge detection information, possible defective areas and the relevant preliminary analysis results; Through the deep learning model in the cloud, feature extraction is performed to generate feature maps; generating a probability distribution based on the feature map, the probability distribution representing a confidence level for each defect type; The classification results are screened according to the preset confidence threshold, the classification results below the threshold are filtered, and the defect classification results are obtained.

[0013] A system for detecting the quality of a cover piece, the system comprising: The image preprocessing module collects images of the cut pieces in the detection area, obtains high-definition images of the cut pieces under standard lighting conditions, and performs preprocessing; The edge recognition module selects edge detection parameters that are suitable for the material and cutting process of the cut piece based on the shape characteristics and edge quality requirements of the face cover and dynamically adjusts the gradient threshold and operator scale to identify the edge of the cut piece based on the adaptive edge detection algorithm. The defect detection module builds an edge defect recognition model, and based on the edge contour data of the cut piece, performs defect recognition and classification on the detected edge contour to obtain the edge defect detection result; The cloud analysis module uploads the cut piece image data and preliminary analysis results to the cloud to obtain the cloud analysis results for the defect types that cannot be determined by edge detection and defect identification; The result generation module calculates the comprehensive quality score of the cut pieces according to the edge defect detection results and the cloud analysis results, and generates the final quality inspection results.

[0014] Furthermore, the cloud analysis module includes: Defect classification unit, based on deep learning models in the cloud, analyzes and classifies the types of defects that cannot be determined; A position determination unit, based on the defect classification data, determines the precise position and boundary of the defect and generates corresponding coordinate data and area size information; A grading and marking unit, which calculates the severity of the defect according to the basic characteristic parameters of the defect and grades and marks the defect; The result output unit obtains corresponding processing instructions and defect identification parameters according to the defect classification data and classification marks, and outputs the cloud analysis results.

[0015] The technical solution of the present invention can achieve the following technical effects: It effectively solves the problems of low efficiency, poor accuracy and incomplete defect identification in traditional cutting piece quality inspection, accurately identifies the edges and defects of cutting pieces, improves the accuracy and consistency of inspection, and reduces manual errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 It is a flow chart of a method for detecting the quality of cover pieces; Figure 2 It is a schematic diagram of the process of preprocessing the high-definition image of the cut piece; Figure 3 It is a flowchart of the illumination correction algorithm; Figure 4 A schematic diagram of the process of constructing a reflectance correction model; Figure 5A schematic diagram of the process of identifying the edge of a cut piece; Figure 6 A schematic diagram of the process for generating a comprehensive edge recognition image; Figure 7 A flowchart for obtaining cloud analysis results; Figure 8 Schematic diagram of the process for analyzing and classifying the defect types that cannot be determined. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0020] Embodiment 1 like Figure 1 As shown, the present invention provides a method for detecting the quality of a cover piece, the method comprising: S1: Capture images of the cut pieces in the detection area, obtain high-definition images of the cut pieces under standard lighting conditions, and perform preprocessing; Specifically, firstly, the image of the cut pieces in the inspection area is collected to ensure that high-definition cut piece images are obtained under standard lighting conditions. Standard lighting conditions refer to the use of uniform and stable light sources to illuminate the cut pieces to avoid image distortion caused by reflections, shadows or uneven lighting. For this reason, diffuse light sources or industrial standard light sources are usually used to ensure constant ambient lighting for image acquisition, thereby improving the usability of the image. During the acquisition process, high-resolution industrial cameras are used to capture the details of the cut pieces. Industrial cameras have high pixel density and usually provide high-definition images that can clearly show the edges and surface details of the cut pieces. In addition, high-precision lenses help eliminate image distortion and ensure consistent clarity of each part of the image, laying the foundation for subsequent image processing.

[0021] S2: According to the shape characteristics and edge quality requirements of the face cover, based on the adaptive edge detection algorithm, select the edge detection parameters that are suitable for the material and cutting process of the cut piece, dynamically adjust the gradient threshold and operator scale, and identify the edge of the cut piece; Specifically, the main purpose of this step is to accurately identify the edge features of the cover piece and ensure that the detected edges meet the shape features and edge quality requirements of the piece. Since the materials of the car seat cover are diverse (such as leather, fabric, imitation leather, etc.), and different cutting processes (such as laser cutting, mold cutting, etc.) will produce different edge characteristics, it is necessary to select edge detection parameters that are suitable for the material and cutting process of the piece. These parameters include gradient threshold and operator scale, which determine how the edge detection algorithm identifies the edge of the piece. The goal of the adaptive edge detection algorithm is to dynamically adjust these parameters so that accurate edge detection results can be obtained under different conditions. For the shape features and edge quality requirements of the piece, it is necessary to ensure that the edge detection results are consistent with the design standards so that in the subsequent defect identification process, possible edge defects (such as burrs, irregularities, breakage, etc.) can be accurately identified. In actual operation, the adaptive edge detection algorithm will dynamically adjust the gradient threshold to improve the detection accuracy by analyzing the local contrast and noise level of the image.

[0022] S3: construct an edge defect recognition model, perform defect recognition and classification on the detected edge contour based on the edge contour data of the cut piece, and obtain the edge defect detection result; Specifically, an edge defect recognition model is constructed based on the edge contour data of the cutting piece to automatically identify and classify defects on the edge of the cutting piece. This model detects possible defects such as edge burrs, cracks, irregular cutting, etc. by analyzing the edge characteristics of the cutting piece, and classifies these defects to generate specific edge defect detection results, which can effectively improve the efficiency and accuracy of detection. Especially in the production process, edge defects are often difficult to identify by the naked eye or simple rule detection. By building an intelligent recognition model, human intervention can be reduced, edge problems can be detected quickly and accurately, and the missed detection rate can be reduced. In addition, the classification function can further help the production line make different processing decisions, avoid serious defects from flowing into subsequent processes, and improve the overall quality of the product and the level of automation in the production process.

[0023] S4: For defect types that cannot be determined by edge detection and defect identification, upload the cut piece image data and preliminary analysis results to the cloud to obtain cloud analysis results; Specifically, the main purpose of this step is to solve complex defects that cannot be determined by local edge detection and defect recognition systems. By uploading the cutting image data and preliminary analysis results to the cloud, and using the cloud's deep learning model for further analysis and classification, cloud-based analysis has stronger computing power and can handle subtle or complex defects that are difficult for local systems to identify, such as small cracks, irregular cutting, etc., thereby improving the comprehensiveness and accuracy of detection. This step also helps optimize resource allocation and reduces the burden on local systems through local and cloud collaboration. In addition, cloud-based analysis can generate more detailed defect classification reports, providing accurate basis for subsequent production decisions (such as repair or scrapping), and ensuring the integrity and intelligence of cutting quality inspection.

[0024] S5: Based on the edge defect detection results and cloud analysis results, the comprehensive quality score of the cut pieces is calculated to generate the final quality inspection results.

[0025] Specifically, first, the defect information obtained from edge defect detection (such as edge burrs, cracks, irregular cutting, etc.) is combined with the detailed defect classification, severity and location data provided by cloud analysis to comprehensively evaluate the overall quality of the cut piece; then, the severity of each defect is weighted and calculated through the pre-set quality scoring criteria. The scoring criteria usually take into account the type, location, size of the defect and the impact on the product function and appearance. For example, larger edge damage or cracks in key areas will result in a lower score, while minor flaws may only have a small impact on the final score; by accumulating the impact of each defect, a comprehensive quality score for the cut piece is generated. According to the scoring results, the system will classify the cut piece into different quality levels (such as qualified, needing repair, unqualified, etc.), and generate a final quality inspection report, which will be provided to the production management system as a basis for whether to rework or continue production.

[0026] The present invention effectively solves the problems of low efficiency, poor accuracy and incomplete defect identification in traditional cut piece quality inspection, accurately identifies the edges and defects of the cut pieces, improves the accuracy and consistency of inspection, and reduces manual errors.

[0027] As a preferred embodiment of the above, Figure 2 As shown, the preprocessing includes: A10: grayscale processing is performed on the cut pieces of different materials to obtain grayscale images; A20: Apply a lighting correction algorithm based on the reflective properties of the cut material to obtain an image with uniform lighting; A30: Use filtering method to remove noise in the image and obtain a denoised image; A40: Use image enhancement technology to improve image contrast and obtain an enhanced image.

[0028] Specifically, the collected images need to be preprocessed to meet the needs of subsequent detection. First, grayscale processing is performed to convert color images into grayscale images, simplify the image structure, reduce unnecessary color information interference, and make edge and defect features more prominent; the next step is illumination correction. Since the surfaces of cut pieces of different materials may have uneven illumination problems, such as brightness differences caused by reflection or shadow, illumination correction adjusts the brightness distribution of the image to make the brightness in the image uniform and reduce the detection error caused by uneven illumination; in order to further improve the image quality, noise removal is also required. Noise may be introduced during the image acquisition process, affecting the accuracy of edge detection. Removing noise in the image through filtering methods (such as Gaussian filtering or median filtering) can effectively improve the smoothness of the image; finally, to ensure that the details in the image are fully reflected, image enhancement technology is used to enhance the contrast of the image, especially in areas with low contrast. Through contrast stretching or gamma correction and other methods, the edges, defects and other features of the cut pieces are made clearer, which is convenient for subsequent edge detection and defect identification.

[0029] As a preferred embodiment of the above, Figure 3 As shown, the illumination correction algorithm includes: B10: Perform histogram analysis on each grayscale image to determine the distribution of the brightness of the grayscale image; B20: Based on the local adaptive histogram equalization method, the brightness of different areas of the image is adjusted; B30: Construct a reflectance correction model and apply multispectral imaging technology to compensate for reflective areas, and use shadow monitoring and interpolation methods to correct shadow parts in the image; B40: Adjust and normalize the global brightness, and set the brightness distribution of the image within a standard range through linear stretching and standardization algorithms; B50: Evaluate the lighting correction effect and calibrate the parameters based on the evaluation results.

[0030] Specifically, first, a histogram analysis is performed on each grayscale image to understand the distribution of pixel brightness in the image and determine whether there are overly bright or overly dark areas; then, a local adaptive histogram equalization method is applied to adjust the brightness of different areas of the image. This step can enhance the details of low-contrast areas and make the brightness distribution more balanced; in step B30, a reflectance correction model is constructed, combined with multispectral imaging technology to compensate for the brightness of the reflective area, and the shadow part in the image is corrected by shadow monitoring and interpolation methods. This process ensures that the brightness difference caused by reflection and shadow in the image is effectively corrected; then, the global brightness of the image is adjusted and normalized, and a linear stretching and standardization algorithm is used to make the brightness value of the image within a predetermined standard range to ensure the brightness consistency of different cut images; finally, the illumination correction effect is evaluated, and the correction parameters are adjusted according to the evaluation results. If there is still unevenness in the image brightness, the parameters will be automatically adjusted to optimize the correction effect. The entire illumination correction process ensures the brightness consistency of the cut image, providing a reliable basis for subsequent edge detection and defect recognition.

[0031] As a preferred embodiment of the above, Figure 4 As shown, constructing the reflectance correction model includes: C10: Obtain the optical properties of each material of the cover piece under different light sources and build a reflection property database; C20: Determine the key parameters of the model and construct a Phong-based illumination model to describe the reflection behavior of the material surface; C30: Based on the illumination model and reflection characteristic database, a reflectivity correction formula is constructed to calculate the brightness compensation value of the reflective area; C40: Use the actual cut piece image data to test the illumination model, calculate the error of the correction effect, and adjust the model parameters to obtain the final reflectivity correction model.

[0032] Specifically, first, the optical properties of various materials of the cover pieces (such as leather, fabric, imitation leather, etc.) under different light source conditions are obtained, including reflectivity, absorptivity, refractive index, etc. By collecting these data, a reflection characteristic database is constructed. The database contains the reflection behavior of different materials under different lighting angles and intensities. This data is crucial for accurately describing the performance of the pieces under lighting conditions in the production environment; then, according to the optical properties of the material, the key parameters required to build the reflectivity correction model are determined. These parameters include the angle of the incident light, the surface roughness of the material, and the light intensity. Based on these parameters, the Phong reflection model is used to describe the reflection behavior of the material surface. The Phong model is a classic lighting model that is suitable for simulating the reflection effects of smooth and rough surfaces. This model can accurately describe the diffuse reflection of the piece material. The reflection characteristics of the reflection, highlight and ambient light are calculated; in step C30, the illumination model and the previously constructed reflection characteristic database are combined to construct a reflectance correction formula, which is used to calculate the brightness compensation value of the reflective area in the image, that is, how to perform brightness correction on the over-reflective area under specific light source conditions. This process can accurately estimate and correct the uneven brightness area in the image according to the incident angle of the light and the reflectance characteristics of the surface of the cut piece, so as to ensure the illumination consistency of the overall image; finally, in step C40, the illumination model is tested using the actual collected cut piece image data, and the accuracy of the model is evaluated by calculating the error after the model correction. If it is found that there is a large difference between the corrected image and the image under ideal lighting conditions, the model parameters will be adjusted according to the error feedback, so as to gradually optimize the model until the final high-precision reflectance correction model is obtained.

[0033] As a preferred embodiment of the above, Figure 5 As shown, step S2, according to the shape characteristics and edge quality requirements of the face cover, based on the adaptive edge detection algorithm, selects edge detection parameters that are suitable for the material and cutting process of the cut piece, dynamically adjusts the gradient threshold and operator scale, and identifies the edge of the cut piece, including: S21: determine the shape characteristics and edge quality requirements of the cut piece, and select edge detection parameters suitable for the cut piece material and cutting process, including an initial gradient threshold and operator scale; S22: applying an adaptive edge detection algorithm to the collected cut piece image data, and dynamically adjusting the gradient threshold according to the local contrast and noise level of the image; S23: Adaptively adjust the scale of the edge detection operator according to the thickness and curvature changes of the edge of the cut piece; S24: Using a multi-scale edge detection algorithm, edge detection results at different scales are integrated to generate a comprehensive edge recognition image; S25: Perform morphological processing on the generated edge recognition image to obtain a final edge recognition result.

[0034] Specifically, first, it is necessary to determine the specific shape characteristics of the cutting piece and the requirements for edge quality. This step is to select edge detection parameters suitable for the cutting piece material (such as leather, fabric, etc.) and cutting process (such as laser cutting, mechanical cutting, etc.), to ensure that the initial gradient threshold and operator scale can match the actual working conditions. The selection of these initial parameters directly affects the effect of edge detection and is therefore the basis of the entire algorithm. After obtaining the high-definition image data of the cutting piece, the adaptive edge detection algorithm is applied to start processing the image data. The algorithm will first analyze the local contrast and noise level of the image, and dynamically adjust the gradient threshold based on these analysis results. The purpose of this is to enable the algorithm to perform targeted detection in different image areas. For example, when the contrast of some areas of the cutting piece image is low, the system will automatically lower the gradient threshold to enhance the ability to capture subtle edges; in areas with more noise, the algorithm will increase the gradient threshold to avoid noise being misjudged as a valid edge; then, the algorithm will adaptively adjust the operator scale of edge detection according to the thickness and curvature changes of the cutting piece edge. For parts with complex edge details, smaller operator scales can capture subtle changes and ensure the accuracy of edge detection. For areas with smoother or thicker edges, using larger operator scales can not only improve detection efficiency, but also reduce computational load and increase overall processing speed. In order to further improve the accuracy of edge detection, a multi-scale edge detection algorithm is used to process the cut-out image at different scales to obtain edge information at different levels. By integrating the edge detection results at different scales, a more complete and accurate edge recognition image can be generated. This multi-scale detection method can effectively cope with the diversity of cut-out edges, so that the algorithm can not only process small edge features, but also accurately identify larger edge structures. Finally, the generated edge recognition image is subjected to morphological processing. Morphological processing is to remove isolated noise points and enhance the coherence and integrity of the edge through operations such as corrosion, expansion, opening and closing operations. Through these morphological operations, the edge detection results become clearer and more stable, and the final high-quality edge recognition results are obtained.

[0035] As a preferred embodiment of the above, Figure 6 As shown, step S24, by using a multi-scale edge detection algorithm, integrates edge detection results at different scales to generate a comprehensive edge recognition image, including: S241: performing multi-scale decomposition on the cut piece image to generate multiple image versions of different scales, and using image decomposition technology to downsample the original image level by level to form image pyramids of different resolutions; S242: applying edge detection algorithms to images of different scales to obtain edge detection results at different scales; S243: Integrate edge detection results at different scales based on a fusion strategy to generate a comprehensive edge recognition image.

[0036] Specifically, first, the cut image is decomposed into multiple scales to generate multiple image versions of different scales. The original image is gradually downsampled through the image pyramid decomposition technology to form multiple image levels with different resolutions. The high-resolution image retains more detail information and is suitable for detecting tiny edges; the low-resolution image pays more attention to the overall structural contour and helps to identify larger or smoother edge areas; next, the edge detection algorithm is applied to each scale image separately. For example, the classic edge detection methods such as Sobel, Canny or Laplacian are used to process the images of each scale separately to obtain edge detection results at different resolutions. The high-resolution image helps to capture delicate edge features, while the low-resolution image can more effectively identify large contour shapes and ensure comprehensive recognition of edge features; then, based on the predetermined fusion strategy, the edge detection results at different scales are integrated to generate a comprehensive edge recognition image. The edge information of different scales can be combined by using methods such as weighted average or maximum response, which not only maintains sensitivity to tiny edges but also ensures the integrity of the global structure. Through this integration step, the generated comprehensive edge recognition image can accurately capture the various levels of complex edge features.

[0037] As a preferred embodiment of the above, Figure 7 As shown, in step S4, for the defect type that cannot be determined by edge detection and defect identification, the cut piece image data and preliminary analysis results are uploaded to the cloud to obtain the cloud analysis results, including: S41: Analyze and classify the unidentified defect types based on the cloud-based deep learning model; S42: Based on the defect classification data, determine the precise location and boundary of the defect, and generate corresponding coordinate data and area size information; S43: Calculate the severity of the defect according to the basic characteristic parameters of the defect, and grade and mark the defect; S44: Obtain corresponding processing instructions and defect identification parameters according to the defect classification data and classification marks, and output cloud analysis results.

[0038] Specifically, first, when local detection cannot identify certain complex or subtle defects, the relevant cutting piece images and preliminary analysis data will be automatically transmitted to the cloud server. The cloud server has stronger computing power and can process large-scale data. It can also apply pre-trained deep learning models to further analyze these unclassifiable defects. In the cloud, the uploaded cutting piece images are first analyzed by a deep learning model. This model is usually based on a convolutional neural network (CNN) or other deep learning frameworks suitable for image analysis. It uses a multi-level feature extraction method to automatically identify defect-related features from the image, such as irregular edges, cracks, burrs, etc. In this way, the deep learning model can analyze the type of defects and classify them into different categories, such as edge burrs, cracks, irregular cutting or other surface defects. Subsequently, the cloud system further determines the precise location and boundaries of each defect based on the defect classification data output by the deep learning model. This is usually done through pixel-level processing of the image. The system can generate coordinate data and area size information of the defect and accurately locate the position of the defect on the cut piece. This step provides specific spatial positioning for subsequent defect processing, ensuring that subsequent repair or rework operations can be concentrated on the defective area. Next, the system will calculate the severity of the defect based on the basic characteristic parameters of the defect (such as the area, depth, position, etc. of the defect) and grade each defect. The severity of the defect is usually evaluated based on its impact on the overall quality of the cut piece. Small or mild defects may not affect the performance of the product and are low-level defects, while larger cracks or edge damage may cause the cut piece to be scrapped and are high-level defects. Through this classification, the system can provide instructions for subsequent production processing; finally, based on the defect classification data and severity classification, the cloud system generates corresponding processing instructions and optimized defect identification parameters. The processing instructions may include repair suggestions, rework instructions or scrap suggestions, and the optimized defect identification parameters can be fed back to the local system to improve the accuracy and efficiency of the subsequent inspection process. The cloud analysis results are finally output, including the defect classification, location, size, severity and processing suggestions, providing a complete basis for further processing and quality assessment of the cut pieces.

[0039] As a preferred embodiment of the above, Figure 6 As shown, step S41, based on the deep learning model in the cloud, analyzes and classifies the types of defects that cannot be determined, including: S411: For defects that cannot be determined, relevant cut piece image data and preliminary analysis results are uploaded to the cloud. The data includes high-definition images of cut pieces, edge detection information, possible defect areas, and relevant preliminary analysis results. S412: extracting features through a deep learning model in the cloud and generating a feature map; S413: Generate a probability distribution based on the feature graph, where the probability distribution represents the confidence level of each defect type; S414: Filter the classification results according to a preset confidence threshold, filter the classification results below the threshold, and obtain the defect classification results.

[0040] Specifically, first, the relevant cutting piece image data and preliminary analysis results are uploaded to the cloud. The data includes high-definition images of the cutting pieces, edge detection information, possible defect areas and preliminary analysis results. These data provide a basis for further analysis on the cloud. Then, the uploaded cutting piece images are processed using the deep learning model on the cloud. The features of the images are extracted through multiple convolutional layers to generate feature maps. The feature maps can capture subtle features and complex defect patterns in the images, thereby helping to identify different defect types. In step S413, based on the generated feature maps, the probability distribution of various types of defects is calculated. The probability distribution represents the confidence of each defect type, that is, the possibility of judging that a certain area belongs to a specific defect. In this way, a corresponding confidence value can be assigned to each defect type. Finally, the classification results are screened according to a preset confidence threshold. Classification results below the threshold will be filtered out to ensure that only results with higher confidence are retained. After screening, the final defect classification results are obtained, which provide an accurate basis for further quality assessment and processing.

[0041] Embodiment 2 Based on the same inventive concept as the method for detecting the quality of a cover piece in the aforementioned embodiment, the present invention further provides a system for detecting the quality of a cover piece, the system comprising: A system for detecting the quality of a cover piece, the system comprising: The image preprocessing module collects images of the cut pieces in the detection area, obtains high-definition images of the cut pieces under standard lighting conditions, and performs preprocessing; The edge recognition module selects edge detection parameters that are suitable for the material and cutting process of the cut piece based on the shape characteristics and edge quality requirements of the face cover and dynamically adjusts the gradient threshold and operator scale to identify the edge of the cut piece based on the adaptive edge detection algorithm. The defect detection module builds an edge defect recognition model, and based on the edge contour data of the cut piece, performs defect recognition and classification on the detected edge contour to obtain the edge defect detection result; The cloud analysis module uploads the cut piece image data and preliminary analysis results to the cloud to obtain the cloud analysis results for the defect types that cannot be determined by edge detection and defect identification; The result generation module calculates the comprehensive quality score of the cut pieces based on the edge defect detection results and cloud analysis results, and generates the final quality inspection results.

[0042] The above-mentioned detection system in the present invention can effectively implement the method for detecting the quality of the cover piece, and the technical effects that can be achieved are as described in the above-mentioned embodiments, which will not be repeated here.

[0043] As a preferred embodiment of the above, the cloud analysis module includes: Defect classification unit, based on deep learning models in the cloud, analyzes and classifies the types of defects that cannot be determined; A position determination unit, based on the defect classification data, determines the precise position and boundary of the defect and generates corresponding coordinate data and area size information; The grading and marking unit calculates the severity of the defect according to the basic characteristic parameters of the defect and grades and marks the defect; The result output unit obtains the corresponding processing instructions and defect identification parameters based on the defect classification data and classification marks, and outputs the cloud analysis results.

[0044] Similarly, the above-mentioned optimization schemes for the system can also respectively achieve the optimization effects corresponding to the method in Example 1, which will not be repeated here.

[0045] Although the present application has been described in conjunction with specific features and embodiments thereof, it is obvious that various modifications and combinations may be made thereto without departing from the spirit and scope of the present application. Accordingly, this specification and the accompanying drawings are merely exemplary illustrations of the present application as defined therein, and are deemed to have covered any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application is intended to include these modifications and variations.

Claims

1. A method for detecting the quality of a cover piece, characterized in that: include: Capture images of the cut pieces in the inspection area, obtain high-definition images of the cut pieces under standard lighting conditions, and perform preprocessing; According to the shape characteristics and edge quality requirements of the face cover, based on the adaptive edge detection algorithm, the edge detection parameters suitable for the material and cutting process of the cut piece are selected, and the gradient threshold and operator scale are dynamically adjusted to identify the edge of the cut piece; Construct an edge defect recognition model, and based on the edge contour data of the cut piece, perform defect recognition and classification on the detected edge contour to obtain the edge defect detection result; For defect types that cannot be determined by edge detection and defect identification, upload the cut piece image data and preliminary analysis results to the cloud to obtain cloud analysis results; Based on the edge defect detection results and cloud analysis results, the comprehensive quality score of the cut pieces is calculated to generate the final quality inspection results.

2. The method for detecting the quality of the cover piece according to claim 1, characterized in that: The pre-processing comprises: Grayscale processing is performed on the cut pieces of different materials to obtain grayscale images; Apply a lighting correction algorithm based on the reflective properties of the cut material to obtain an image with uniform lighting; A filtering method is used to remove noise from the image to obtain a denoised image; Image enhancement technology is used to improve image contrast and obtain an enhanced image.

3. The method for detecting the quality of the cover piece according to claim 2, characterized in that: The illumination correction algorithm includes: Performing a histogram analysis on each of the grayscale images to determine the distribution of the brightness of the grayscale images; Based on the local adaptive histogram equalization method, the brightness of different areas of the image is adjusted; Construct a reflectivity correction model and apply multispectral imaging technology to compensate for reflective areas, and use shadow monitoring and interpolation methods to correct shadow parts in the image; Adjust and normalize the global brightness, and set the brightness distribution of the image within a standard range through linear stretching and standardization algorithms; The lighting correction effect is evaluated and the parameters are corrected based on the evaluation results.

4. The method for detecting the quality of the cover piece according to claim 3, characterized in that: Constructing the reflectivity correction model includes: Obtain the optical properties of each material of the cover piece under different light sources and build a reflection property database; Determine the key parameters of the model and construct a Phong-based illumination model to describe the reflection behavior of the material surface; Based on the illumination model and the reflection characteristic database, a reflectivity correction formula is constructed to calculate the brightness compensation value of the reflective area; The illumination model is tested using actual cut piece image data, the error of the correction effect is calculated, and the model parameters are adjusted to obtain the final reflectivity correction model.

5. The method for detecting the quality of cover pieces according to claim 1, characterized in that: According to the shape characteristics and edge quality requirements of the face cover, based on the adaptive edge detection algorithm, select the edge detection parameters that are suitable for the material and cutting process of the cut piece, dynamically adjust the gradient threshold and operator scale, and identify the edge of the cut piece, including: Determine the shape characteristics and edge quality requirements of the cut piece, and select edge detection parameters suitable for the cut piece material and cutting process, including the initial gradient threshold and operator scale; Applying an adaptive edge detection algorithm to the collected image data of the cut piece, dynamically adjusting the gradient threshold according to the local contrast and noise level of the image; Adaptively adjust the scale of the edge detection operator according to the thickness and curvature changes of the edge of the cut piece; Through the multi-scale edge detection algorithm, the edge detection results at different scales are integrated to generate a comprehensive edge recognition image; The generated edge recognition image is subjected to morphological processing to obtain the final edge recognition result.

6. The method for detecting the quality of the cover piece according to claim 5, characterized in that: Through the multi-scale edge detection algorithm, the edge detection results at different scales are integrated to generate a comprehensive edge recognition image, including: Perform multi-scale decomposition on the cut image to generate multiple image versions of different scales. Use image decomposition technology to downsample the original image step by step to form image pyramids of different resolutions. Applying edge detection algorithms to the images of each scale respectively to obtain edge detection results at different scales; The edge detection results at different scales are integrated based on the fusion strategy to generate a comprehensive edge recognition image.

7. The method for detecting the quality of cover pieces according to claim 1, characterized in that: For defect types that cannot be determined by edge detection and defect identification, upload the cut piece image data and preliminary analysis results to the cloud to obtain cloud analysis results, including: Analyze and classify the unidentified defect types based on the cloud-based deep learning model; Based on the defect classification data, determine the exact location and boundary of the defect, and generate the corresponding coordinate data and area size information; Calculate the severity of the defect based on the basic characteristic parameters of the defect and grade and mark the defect; According to the defect classification data and grading marks, corresponding processing instructions and defect identification parameters are obtained, and cloud analysis results are output.

8. The method for detecting the quality of cover pieces according to claim 1, characterized in that: The cloud-based deep learning model analyzes and classifies the types of defects that cannot be determined, including: If the defect type cannot be determined, the relevant image data of the cut piece and the preliminary analysis results are uploaded to the cloud, including the high-definition image of the cut piece, edge detection information, possible defective areas and the relevant preliminary analysis results; Through the deep learning model in the cloud, feature extraction is performed to generate feature maps; generating a probability distribution based on the feature map, the probability distribution representing a confidence level for each defect type; The classification results are screened according to the preset confidence threshold, the classification results below the threshold are filtered, and the defect classification results are obtained.

9. A system for detecting the quality of cover pieces, characterized in that: The system comprises: The image preprocessing module collects images of the cut pieces in the detection area, obtains high-definition images of the cut pieces under standard lighting conditions, and performs preprocessing; The edge recognition module selects edge detection parameters that are suitable for the material and cutting process of the cut piece based on the shape characteristics and edge quality requirements of the face cover and dynamically adjusts the gradient threshold and operator scale to identify the edge of the cut piece based on the adaptive edge detection algorithm. The defect detection module builds an edge defect recognition model, and based on the edge contour data of the cut piece, performs defect recognition and classification on the detected edge contour to obtain the edge defect detection result; The cloud analysis module uploads the cut piece image data and preliminary analysis results to the cloud to obtain the cloud analysis results for the defect types that cannot be determined by edge detection and defect identification; The result generation module calculates the comprehensive quality score of the cut pieces according to the edge defect detection results and the cloud analysis results, and generates the final quality inspection results.

10. The cover piece quality inspection system according to claim 9, characterized in that: The cloud analysis module includes: Defect classification unit, based on deep learning models in the cloud, analyzes and classifies the types of defects that cannot be determined; A position determination unit, based on the defect classification data, determines the precise position and boundary of the defect and generates corresponding coordinate data and area size information; A grading and marking unit, which calculates the severity of the defect according to the basic characteristic parameters of the defect and grades and marks the defect; The result output unit obtains corresponding processing instructions and defect identification parameters according to the defect classification data and classification marks, and outputs the cloud analysis results.

Images