Grey cloth pattern defect visibility prediction method and system

Abstract

The invention relates to the technical field of image processing, and discloses a gray fabric pattern flaw visibility prediction method and system, and the method comprises the steps of data collection, data preprocessing, feature extraction, feature enhancement, model training, model verification, flaw visibility prediction and system implementation. According to the gray fabric pattern flaw visibility prediction method and system, through tip image processing and machine learning, a major breakthrough of gray fabric flaw detection is achieved, flaws are captured with unprecedented precision, flaw visibility is intelligently predicted, the missed detection risk is greatly reduced, accuracy is improved, compared with manual detection and automatic prediction, the speed is remarkably increased, and the detection efficiency is improved. Labor cost is reduced, personal errors are reduced, stability and reliability are improved, flaws are recognized in an early stage, reworking and scrapping are reduced, 24-hour work of an automatic detection system is ensured, high efficiency and consistency are ensured, operation cost is reduced, and market competitiveness and profitability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method and system for predicting visibility of grey cloth pattern defects. Background Art

[0002] In the field of industrial manufacturing, surface defect detection technology is a key method to obtain material surface information and perform quality analysis through optical imaging systems.

[0003] The process of attaching images to materials includes industrial processes such as printing, weaving, and laminating. However, the existence of defects in the production process is still an unavoidable problem. Traditional manual defect detection methods are not only inefficient, but also have the risk of false detection and missed detection. With the development of machine vision technology, automated defect detection systems are gradually applied to industrialization, but there is still the problem of inaccurate defect visibility prediction, especially in the image acquisition stage. Some defects are difficult to be captured by the camera because they are not obvious. Therefore, there is an urgent need for defect visibility prediction methods and systems. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for predicting the visibility of grey fabric pattern defects to solve the problems raised in the above background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for predicting the visibility of defects in grey fabric patterns, comprising the following steps: S1: data acquisition, wherein a data acquisition module acquires grey fabric images from a production line, the images including normal grey fabric images and defective grey fabric images; S2: Data preprocessing: preprocessing the collected image data, including denoising, contrast enhancement, and brightness adjustment to improve image quality. At the same time, the grey cloth image data is cleaned to remove outliers and missing values. S3: Feature extraction, extracting defect features from the preprocessed image data, including the size, shape, color, and texture of the defect, and converting the grey fabric image data into numerical features; S4: Feature enhancement: The extracted features are enhanced to improve the discrimination and stability of the features. The features can be enhanced by calculating the statistics of the features. S5: Model training: Use the processed feature data and corresponding defect visibility labels to train the decision tree model, adjust the model's hyperparameters, and optimize the model's performance. S6: Model validation, use the validation set to validate the trained model and evaluate the model's prediction accuracy and generalization ability; S7: Defect visibility prediction: the image data of the grey cloth to be predicted and the production parameter data are input into the trained decision tree model to obtain the prediction result of defect visibility.

[0006] As a preferred embodiment, the data collection step includes collecting grey cloth image data from the production line and recording the production parameters of the grey cloth, the grey cloth image including the organizational structure and plain weave, twill, satin, warp yarn count, weft yarn count, warp yarn density, and weft yarn density.

[0007] As a preferred embodiment, the data preprocessing step includes denoising, contrast enhancement and brightness adjustment of the image data, and cleaning of the production parameter data.

[0008] As a preferred embodiment, the feature extraction step includes extracting defect features from the preprocessed image data and converting the production parameter data into numerical features.

[0009] As a preferred embodiment, the feature enhancement step includes performing statistical calculation or feature transformation processing on the extracted features to enhance the features, the statistics include mean, variance, and extreme value, and the feature transformation includes principal component analysis and linear discriminant analysis.

[0010] As a preferred implementation method, hyperparameters include the maximum depth of the decision tree, evaluation criteria, and partitioning principles. In addition to using an independent validation set for evaluation, stratified cross-validation is introduced to ensure that the training set and validation set are evenly divided according to defect categories. To verify the adaptability to complex scenarios, an adversarial test set containing dynamic lighting, mechanical vibration blur, and unknown defects is constructed, and actual production line verification is carried out in a cooperative factory.

[0011] As a preferred embodiment, the defect visibility prediction step includes inputting the image data of the grey cloth to be predicted and the production parameter data into the trained decision tree model to obtain the prediction result of the defect visibility. According to the prediction result, it can be judged whether the defect is obvious and whether it is easy to be captured by the camera, thereby guiding the subsequent defect detection and processing work.

[0012] The grey cloth pattern defect visibility prediction system includes a data acquisition module, a data preprocessing module, a feature extraction module, a feature enhancement module, a model training module, a model verification module and a result prediction module, and is used to implement the above-mentioned grey cloth pattern defect visibility prediction method.

[0013] Compared with the prior art, the present invention has the following beneficial effects: First, the present invention has achieved a major breakthrough in the field of grey fabric defect detection by comprehensively applying cutting-edge image processing and machine learning technologies. It can extract the subtle features of various defects on grey fabric with unprecedented accuracy. These features include but are not limited to color differences, texture changes, shape irregularities, and size deviations. Through deep learning and pattern recognition algorithms, the system can intelligently analyze these features and make highly accurate predictions of the visibility of defects based on the results of big data analysis. This innovative method greatly reduces the risk of missed detection due to subtle or hidden defects. In traditional detection methods, manual inspectors may find it difficult to detect certain subtle defects due to fatigue, limited vision, or lack of experience, while the present invention can capture these subtle defects. Subtle differences ensure that every potential defect will not be missed. In addition, the present invention significantly improves the accuracy of defect detection. It can not only identify the existence of defects, but also accurately judge the type, location and severity of the defects. This detailed information provides valuable guidance for subsequent repairs or treatments, which helps companies to more effectively control product quality. From a broader perspective, the application of the present invention is of great significance to improving the production efficiency and product quality of the entire textile industry. By reducing missed inspections and misjudgments, companies can more effectively control production costs while ensuring that the final product meets or even exceeds customer expectations. This not only helps to enhance the company's market competitiveness, but also promotes the entire industry to develop a higher quality and more efficient production model.

[0014] Second, the traditional defect detection method of the present invention has long relied on careful manual observation and judgment. This method is not only time-consuming and labor-intensive, requiring a large amount of human resources, but also has low detection efficiency and is difficult to meet the modern textile industry's demand for high output. In addition, manual detection is easily affected by multiple human factors such as the inspector's experience, concentration, and fatigue, resulting in difficulty in ensuring the accuracy and consistency of the detection results. In contrast, the present invention has achieved a revolutionary change in the detection process by introducing technology for automatically predicting defect visibility. By integrating advanced image processing and machine learning algorithms, the system can autonomously and quickly analyze the grey fabric image and accurately predict the visibility of defects, thereby significantly accelerating the overall detection speed. This automated detection method not only greatly reduces the need for manual intervention, It reduces labor costs and greatly improves production efficiency, allowing companies to complete more inspection tasks in a shorter time. More importantly, the automated inspection method effectively reduces errors caused by human factors. The system is not affected by fatigue, emotions or experience differences, and can always maintain a high degree of consistency and accuracy. This not only improves the reliability of the inspection results, but also helps companies better control product quality and reduce rework and scrap costs caused by defects. In summary, the automated defect detection technology of the present invention not only significantly improves production efficiency and reduces labor costs, but also improves the stability and reliability of the overall production line by reducing human errors. This is undoubtedly a huge blessing for modern textile companies, and helps companies maintain their leading position in the fierce market competition and achieve sustainable development.

[0015] Third, since the present invention has the ability to predict defect visibility with high precision, it can keenly capture potential quality problems on the grey cloth at a very early stage of the production process. This means that once the defect is identified by the system, the company can take immediate action, including adjusting production parameters, replacing raw materials or timely shutting down for maintenance, thereby effectively eliminating the defect before it expands or affects the final product. This approach greatly reduces the rework and scrap caused by defective products flowing into subsequent processes, not only avoiding additional repair costs, but also reducing brand reputation losses caused by unqualified products. Furthermore, through the introduction of automated detection technology, the present invention significantly reduces dependence on manual labor. Traditional manual detection requires a large amount of manpower input, which is not only costly, but also often difficult to achieve in terms of efficiency and quality when faced with large-scale, high-intensity production tasks. Guaranteed, while the automated inspection system can work 24 hours a day without interruption, without rest, and is not affected by the working environment or emotional fluctuations, ensuring the consistency and efficiency of inspection. This not only reduces the direct costs incurred by the company for hiring a large number of inspectors, including wages, benefits and training costs, but also indirectly reduces the indirect costs that may be caused by poor human resource management, including employee turnover, recruitment and training cycles. Therefore, from the perspective of production costs, the automated defect detection system of the present invention not only reduces direct material and rework costs by reducing defective products, but also significantly reduces the company's operating costs by reducing manpower input. This dual benefit enables the company to have stronger cost control capabilities while maintaining high-quality output, thereby occupying a more advantageous position in market competition. In the long run, this will help the company achieve higher profitability and more sustainable development. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flow chart of the overall steps of the present invention; Figure 2 This is a flow chart of the data collection steps of the present invention; Figure 3 Flowchart of data preprocessing steps of the present invention; Figure 4 This is a flow chart of the feature extraction steps of the present invention; Figure 5 A flowchart of the steps for enhancing the features of the present invention; Figure 6 This is a flow chart of the model training steps of the present invention; Figure 7 This is a flow chart of the defect visibility prediction steps of the present invention. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention. Example

[0018] include Figure 1 、 Figure 2 、 Figure 3 、 Figure 4 、 Figure 5 、 Figure 6 and Figure 7 As shown, the present invention provides a technical solution: a method and system for predicting the visibility of grey fabric pattern defects, comprising the following steps: S1: data acquisition, a data acquisition module collects grey fabric images from a production line, the images including normal grey fabric images and defective grey fabric images; S2: Data preprocessing: preprocessing the collected image data, including denoising, contrast enhancement, and brightness adjustment to improve image quality. At the same time, the grey cloth image data is cleaned to remove outliers and missing values. S3: Feature extraction, extracting defect features from the preprocessed image data, including the size, shape, color, and texture of the defect, and converting the grey fabric image data into numerical features; S4: Feature enhancement: The extracted features are enhanced to improve the discrimination and stability of the features. The features can be enhanced by calculating the statistics of the features. S5: Model training: Use the processed feature data and corresponding defect visibility labels to train a decision tree model, adjust the model's hyperparameters, and optimize the model's performance. The defect visibility label includes whether the defect is obvious and whether it is easily captured by the camera. S6: Model validation, use the validation set to validate the trained model and evaluate the model's prediction accuracy and generalization ability; S7: Defect visibility prediction: the image data of the grey cloth to be predicted and the production parameter data are input into the trained decision tree model to obtain the prediction result of defect visibility.

[0019] The data collection step includes collecting grey cloth image data from the production line and recording the production parameters of the grey cloth. The grey cloth image includes the weave structure, warp yarn count, weft yarn count, warp yarn density, and weft yarn density; the weave structure includes plain weave, twill, and satin.

[0020] Collecting grey fabric image data from the production line enables the system to directly obtain detailed information about the grey fabric surface, and using advanced image processing and machine learning techniques to conduct in-depth analysis of the image, thereby accurately identifying any subtle defects. At the same time, recording the production parameters of the grey fabric, including warp and weft density, tension, and speed, provides the system with comprehensive production background information, which helps to more deeply understand the possible causes of defects. Through this comprehensive data collection method, the present invention not only realizes real-time monitoring and accurate evaluation of grey fabric quality, but also provides enterprises with valuable production data support, which helps to optimize production processes, reduce defects, and ultimately reduce production costs, thereby improving overall production efficiency and product quality.

[0021] The data preprocessing steps include denoising the image data, enhancing the contrast and adjusting the brightness, as well as cleaning the production parameter data.

[0022] Denoising, contrast enhancement and brightness adjustment of image data can significantly improve image quality, make defect features more obvious, and lay a solid foundation for subsequent defect identification and analysis. Denoising can reduce noise interference in the image and improve image clarity; enhancing contrast and adjusting brightness can improve the visual effect of the image and make the boundary between defects and normal areas clearer. At the same time, cleaning the production parameter data can remove duplicate, erroneous or inconsistent data, ensure the accuracy and consistency of the data, and provide reliable support for subsequent modeling and analysis. Through this comprehensive data preprocessing step, the present invention not only improves the accuracy of defect detection, but also enhances the stability and reliability of the system, providing enterprises with a more efficient and accurate production quality inspection method.

[0023] The feature extraction step includes extracting defect features from the preprocessed image data and converting the production parameter data into numerical features.

[0024] This step extracts defect features from the preprocessed image data. These features include but are not limited to the shape, size, color, and texture of the defects, which are important bases for subsequent defect identification and classification. At the same time, the feature extraction step also converts production parameter data into numerical features. This process allows the production parameters that may originally be complex and diverse to be quantified, facilitating subsequent modeling and analysis. Through this comprehensive feature extraction step, the present invention can more comprehensively capture the quality information of the grey cloth and provide a more accurate and reliable basis for defect detection. This not only improves the accuracy and efficiency of defect detection, but also helps enterprises better control production costs and quality, thereby enhancing overall competitiveness.

[0025] The feature enhancement step involves performing statistical calculations or feature transformations on the extracted features to enhance the features. Statistics include mean, variance, and extreme values, and feature transformations include principal component analysis and linear discriminant analysis.

[0026] Specifically, this step calculates statistics or transforms features on the extracted features, such as calculating the statistics of the mean, standard deviation, maximum, and minimum values of the features. The autoencoder based on deep learning can automatically extract high-level abstract features through unsupervised learning, overcoming the traditional method's reliance on artificial feature design. The generative adversarial network can synthesize a variety of defect samples to expand the training data and enhance the model's ability to recognize rare defects. Customized enhancement algorithms for textile textures such as multi-scale Gabor filtering are composed of Gaussian envelope functions multiplied by complex sinusoidal carriers: , where is the rotating coordinate system, controls the filter frequency domain center frequency, determines the texture period, matches the fabric texture base frequency, is the filter direction angle, covering 0° to 180°, is the Gaussian window scale, determines the filter bandwidth, is the aspect ratio, is the Gaussian window ellipticity, is the phase offset, and multi-scale Gabor. OR filtering simulates the multi-channel characteristics of the human visual system, constructs a filter group at multiple scales and directions, and extracts the frequency domain response of the fabric texture. When the weft shrinkage defect is filtered in a specific direction, it will show a high-energy response, while the periodic texture of the plain weave will show a uniform frequency domain distribution. By fusing Gabor feature maps of different scales, it can capture both tiny defects and macro texture anomalies, focus on the unique periodic texture features of the fabric, and effectively distinguish the subtle differences between normal tissue and defective areas. In addition, the introduction of attention mechanism or graph convolutional network can model the long-range dependency between features and capture the complex interaction patterns of defects and background. These technologies can complement the classical methods by automatically generating more discriminative feature representations, reducing redundant information interference, and adapting to the dynamic changes of different lighting, fabric types and defect shapes, thereby achieving more robust defect visibility prediction in complex industrial scenarios.

[0027] The model training steps include using processed feature data and corresponding defect visibility labels to train a decision tree model, and adjusting the model's hyperparameters to optimize performance. The hyperparameters include the maximum depth of the decision tree, evaluation criteria, and partitioning principles. In addition to using an independent validation set for evaluation, stratified cross-validation is introduced to ensure that the training set and validation set are evenly divided according to defect categories. To verify adaptability to complex scenarios, an adversarial test set containing dynamic lighting, mechanical vibration blur, and unknown defects is constructed, and actual production line verification is carried out in a cooperative factory.

[0028] This step uses the processed feature data and the corresponding defect visibility labels to train the decision tree model, so that the model can learn the complex relationship between features and defect visibility. By continuously adjusting the model's hyperparameters, such as the depth of the tree and the minimum number of samples required for splitting, the model's performance is optimized to ensure its generalization ability and prediction accuracy on new data. This process not only improves the accuracy and efficiency of defect detection, but also helps companies better control production costs and quality risks. The low hardware cost and transparent decision logic of the decision tree are more in line with the rigid requirements of industrial production lines for high speed, low cost, and traceability. Through the model training steps described in the present invention, companies can establish a set of automated defect detection systems based on decision trees to achieve intelligent monitoring and management of grey cloth quality.

[0029] The defect visibility prediction step includes inputting the image data of the grey fabric to be predicted and the production parameter data into the trained decision tree model to obtain the prediction results of the defect visibility. Based on the prediction results, it can be determined whether the defects are obvious and whether they are easy to be captured by the camera, thereby guiding subsequent defect detection and processing work.

[0030] A seamless conversion from data input to prediction output is achieved. This step inputs the grey cloth image data and production parameter data to be predicted into the trained decision tree model. Based on these input data, the model uses the complex relationship between the learned features and defect visibility to quickly and accurately calculate the prediction results of defect visibility. This step not only greatly improves the efficiency and accuracy of defect detection, but also enables enterprises to monitor the quality of grey cloth in real time and automatically on the production line, and promptly discover and deal with potential defect problems. Through the defect visibility prediction step described in the present invention, enterprises can significantly improve product quality and production efficiency, reduce rework and scrap costs caused by defects, and thus occupy an advantageous position in the fierce market competition.

[0031] A grey cloth pattern defect visibility prediction system is implemented by using a data acquisition module, a data preprocessing module, a feature extraction module, a feature enhancement module, a model training module, a model verification module and a result prediction module in the data collection steps, data preprocessing, feature extraction, feature enhancement, model training, model verification and defect visibility prediction.

[0032] The system consists of a data acquisition module, a data preprocessing module, a feature extraction module, a feature enhancement module, a model training module, a model validation module, and a result prediction module, forming a complete process from data collection to prediction output. The data acquisition module is responsible for collecting real-time images of grey fabrics and production parameters from the production line, providing a basis for subsequent analysis. The data preprocessing module improves image quality through denoising, contrast enhancement, and brightness adjustment, while also cleaning production parameter data to ensure data accuracy and consistency. The feature extraction module and feature enhancement module are responsible for extracting key defect features from the preprocessed data and further optimizing these features through statistical calculation or feature transformation to make them more representative and discriminative. The model training module uses these optimized feature data and defect visibility labels to train a decision tree model. The model validation module adjusts hyperparameters using grid search to optimize performance. Finally, the result prediction module inputs the grey fabric data to be predicted into the trained model and outputs the defect visibility prediction results in real time. This complete system not only greatly improves the accuracy and efficiency of defect detection, but also realizes intelligent monitoring of production quality, bringing significant cost savings and competitiveness to enterprises.

[0033] Working Principle: By comprehensively utilizing cutting-edge image processing and machine learning technologies, this invention has achieved a major breakthrough in the field of grey fabric defect detection. It can accurately extract the subtle features of various defects on grey fabric, including but not limited to color differences, texture changes, shape irregularities, and size deviations. Through deep learning and pattern recognition algorithms, the system can intelligently analyze these features and make highly accurate predictions of the visibility of defects based on the results of big data analysis. This innovative method effectively reduces the risk of missed detection due to subtle or hidden defects. In traditional detection methods, manual inspectors may find it difficult to detect certain subtle defects due to fatigue, limited vision, or lack of experience. However, this invention can capture these subtle defects. Differences ensure that no potential defects are missed. In addition, the present invention significantly improves the accuracy of defect detection. It can not only identify the existence of defects, but also accurately judge the type, location and severity of the defects. This detailed information provides valuable guidance for subsequent repairs or treatments, helping companies to more effectively control product quality. From a broader perspective, the application of the present invention is of great significance to improving the production efficiency and product quality of the entire textile industry. By reducing missed detections and misjudgments, companies can more effectively control production costs while ensuring that the final products meet or even exceed customer expectations. This not only helps to enhance the market competitiveness of companies, but also promotes the entire industry to develop a higher quality and more efficient production model. In addition, the traditional defect detection method of the present invention has long relied on careful manual observation and judgment. This method is not only time-consuming and labor-intensive, requiring a large amount of human resources, but also has low detection efficiency and is difficult to meet the modern textile industry's demand for high output. In addition, manual detection is easily affected by multiple human factors such as the inspector's experience, concentration, and fatigue, resulting in difficulty in ensuring the accuracy and consistency of the detection results. In contrast, the present invention has achieved a revolutionary change in the detection process by introducing technology for automatically predicting defect visibility. By integrating advanced image processing and machine learning algorithms, the system can autonomously and quickly analyze the grey fabric image and accurately predict the visibility of defects, thereby significantly accelerating the overall detection speed. This automated detection method not only greatly reduces the need for human intervention and reduces labor costs, but also significantly improves production efficiency, allowing companies to complete more detection tasks in a shorter time. More importantly, the automated detection method effectively reduces errors caused by human factors. The system is not affected by fatigue, emotions or experience differences and can always maintain a high degree of consistency and accuracy. This not only improves the reliability of the detection results, but also helps companies better control product quality and reduce rework and scrap costs caused by defects. In summary, the automated defect detection technology of the present invention not only significantly improves production efficiency and reduces labor costs, but also improves the stability and reliability of the overall production line by reducing human errors. Reliability is undoubtedly a great blessing for modern textile companies, helping them maintain their leading position in the fierce market competition and achieve sustainable development. Moreover, it has high-precision defect visibility prediction capabilities, which can keenly capture potential quality problems on the grey cloth at a very early stage of the production process. This means that once the defect is identified by the system, the company can take immediate action, including adjusting production parameters, replacing raw materials or shutting down for maintenance in time, thereby effectively eliminating the defect before it expands or affects the final product. This approach greatly reduces the rework and scrap caused by defective products flowing into subsequent processes, which not only avoids additional repair costs, but also reduces the loss of brand reputation due to unqualified products. Furthermore, through the introduction of automated detection technology, the present invention significantly reduces dependence on manual labor. Traditional manual detection requires a large amount of manpower input, which is not only costly, but also difficult to guarantee the efficiency and quality of manual detection when facing large-scale, high-intensity production tasks. The automated detection system can work 24 hours a day without interruption, without rest, and is not affected by the working environment or emotional fluctuations, ensuring the consistency and efficiency of detection. This not only reduces the direct costs incurred by enterprises for hiring a large number of detection personnel, including wages, benefits and training costs, but also indirectly reduces the indirect costs that may be caused by poor human resource management, including employee turnover, recruitment and training cycles. Therefore, from the perspective of production costs,The automated defect detection system of this invention not only reduces direct material and rework costs by reducing defective products, but also significantly lowers the company's operating costs by reducing manpower input. This dual benefit enables companies to maintain high-quality output while having stronger cost control capabilities, thereby gaining a more advantageous position in the market competition. In the long run, this will help companies achieve higher profitability and more sustainable development.

[0034] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and alterations may be made to the embodiments without departing from the principles and spirit thereof, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for predicting the visibility of grey fabric pattern defects, characterized by: The following steps are involved: S1: Data acquisition: The data acquisition module collects images of grey fabrics from the production line. The images include normal grey fabric images and defective grey fabric images. S2: Data preprocessing: preprocessing the collected image data, including denoising, contrast enhancement, and brightness adjustment to improve image quality. At the same time, the grey cloth image data is cleaned to remove outliers and missing values. S3: Feature extraction, extracting defect features from the preprocessed image data, including the size, shape, color, and texture of the defect, and converting the grey fabric image data into numerical features; S4: Feature enhancement: The extracted features are enhanced to improve the discrimination and stability of the features. The features can be enhanced by calculating the statistics of the features. S5: Model training: Use the processed feature data and corresponding defect visibility labels to train the decision tree model, adjust the model's hyperparameters, and optimize the model's performance. S6: Model validation, use the validation set to validate the trained model and evaluate the model's prediction accuracy and generalization ability; S7: Defect visibility prediction: the image data of the grey cloth to be predicted and the production parameter data are input into the trained decision tree model to obtain the prediction result of defect visibility.

2. The method for predicting the visibility of grey fabric pattern defects according to claim 1, wherein: The data collection step includes collecting grey cloth image data from the production line and recording the production parameters of the grey cloth. The grey cloth image includes the weave structure, warp yarn count, weft yarn count, warp yarn density, and weft yarn density.

3. The method for predicting the visibility of grey fabric pattern defects according to claim 1, wherein: The data preprocessing steps include denoising the image data, enhancing the contrast and adjusting the brightness, as well as cleaning the production parameter data.

4. The method for predicting the visibility of grey fabric pattern defects according to claim 1, wherein: The feature extraction step includes extracting defect features from the preprocessed image data and converting the production parameter data into numerical features.

5. The method for predicting the visibility of grey fabric pattern defects according to claim 1, wherein: The feature enhancement step includes performing statistical calculations or feature transformation processing on the extracted features to enhance the features. The statistics include mean, variance, and extreme values. The feature transformation includes principal component analysis and linear discriminant analysis.

6. The method for predicting the visibility of grey fabric pattern defects according to claim 1, wherein: Hyperparameters include the maximum depth of the decision tree, evaluation criteria, and partitioning principles. In addition to using an independent validation set for evaluation, stratified cross-validation is introduced to ensure that the training set and validation set are evenly divided according to defect categories. To verify adaptability to complex scenarios, an adversarial test set containing dynamic lighting, mechanical vibration blur, and unknown defects is constructed, and actual production line verification is carried out in a cooperative factory.

7. The method for predicting the visibility of grey fabric pattern defects according to claim 1, wherein: The defect visibility prediction step includes inputting the image data of the grey fabric to be predicted and the production parameter data into the trained decision tree model to obtain the prediction results of the defect visibility. Based on the prediction results, it can be determined whether the defects are obvious and whether they are easy to be captured by the camera, thereby guiding subsequent defect detection and processing work.

8. The system for predicting the visibility of grey fabric pattern defects according to claim 1, characterized in that: The method comprises a data acquisition module, a data preprocessing module, a feature extraction module, a feature enhancement module, a model training module, a model verification module and a result prediction module, and is used to implement the grey cloth pattern defect visibility prediction method according to any one of claims 1 to 7.

Images