Sanded woven fabric quality evaluation method and system combined with machine vision

Abstract

The invention provides a sanded woven fabric quality evaluation method and system in combination with machine vision, and relates to the technical field of sanded woven fabrics, and the method comprises the steps: introducing a fabric quality evaluation multi-dimensional model, carrying out the multi-dimensional evaluation of the fabric quality in combination with a fabric production demand and a fabric production image, and constructing a fabric quality evaluation map; carrying out self-adaptive adjustment on the fabric production control scheme; performing fabric quality evaluation optimization according to the fabric quality evaluation multi-dimensional model; guiding the first fabric production regulation domain to perform multiple propagation optimization according to the first regulation guiding population; and performing optimization control on the sueding machine according to a production control optimization result. According to the method and the device, the technical problem that the fabric quality stability is influenced due to insufficient quality evaluation accuracy caused by lack of comprehensive evaluation of multiple dimensions in fabric quality evaluation in the prior art is solved, self-adaptive adjustment of a sueding machine production control scheme is realized through the fabric quality evaluation multi-dimensional model, and the fabric quality evaluation accuracy is improved. And the stability of the fabric quality is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of brushed woven fabric, and particularly relates to a brushed woven fabric quality evaluation method and system combining machine vision. BACKGROUND

[0002] In the traditional fabric production process, especially in the operation of the brushing machine, the fabric quality evaluation usually relies on a single-dimensional detection method, only focusing on the appearance quality, mechanical properties or comfort of the fabric, and lacks comprehensive evaluation of multiple performance dimensions of the fabric, which cannot fully reflect the performance of the fabric in actual use, and cannot achieve precise quality control when facing different production demands, thereby affecting the consistency and stability of the final product. Since the single-dimensional quality evaluation cannot identify the specific dimension of the problem, it cannot be targeted for optimization when adjusting the production parameters, resulting in fluctuations in product quality between different production batches, or even between the same batch, and lack of consistency and stability.

[0003] In summary, the prior art has the technical problem that the fabric quality evaluation relies on single-dimensional detection, lacks comprehensive evaluation of multiple dimensions, resulting in insufficient quality evaluation accuracy, and thereby affecting the stability of the fabric quality. SUMMARY

[0004] The purpose of the present application is to provide a brushed woven fabric quality evaluation method and system combining machine vision, to solve the technical problem in the prior art that the fabric quality evaluation relies on single-dimensional detection, lacks comprehensive evaluation of multiple dimensions, resulting in insufficient quality evaluation accuracy, and thereby affecting the stability of the fabric quality.

[0005] In view of the above problems, the present application provides a brushed woven fabric quality evaluation method and system combining machine vision.

[0006] In a first aspect, the application provides a method for evaluating the quality of a sanding woven fabric by combining machine vision, which is implemented by a sanding woven fabric quality evaluation system combining machine vision. The method includes the following steps: when a sanding machine executes a fabric production instruction, a fabric production image is synchronously acquired, the fabric production instruction including a fabric production control scheme corresponding to a fabric production requirement; a fabric quality evaluation multi-dimensional model is introduced, and a fabric quality multi-dimensional evaluation is performed in combination with the fabric production requirement and the fabric production image to construct a fabric quality evaluation atlas; the fabric production control scheme is adaptively adjusted according to the fabric quality evaluation atlas to obtain a first fabric production adjustment domain; the fabric quality evaluation multi-dimensional model is used to perform fabric quality evaluation optimization on the first fabric production adjustment domain to obtain a first adjustment guide population; based on the fabric quality evaluation multi-dimensional model, the first adjustment guide population is used to guide the first fabric production adjustment domain to perform multiple reproduction optimizations to obtain a production control optimization result; and the sanding machine is controlled based on the production control optimization result.

[0007] Optionally, when the sanding machine executes the fabric production instruction, a fabric production monitoring image is synchronously acquired according to a machine vision device; and the fabric production monitoring image is subjected to denoising processing to obtain the fabric production image.

[0008] Optionally, a multi-dimensional defect is captured from the fabric production image according to the fabric production requirement to obtain a fabric surface defect grid, a fabric texture defect grid and a fabric color defect grid; the fabric quality evaluation multi-dimensional model is activated, the fabric quality evaluation multi-dimensional model including a surface defect evaluation model, a texture defect evaluation model and a color defect evaluation model; the fabric surface defect grid is input into the surface defect evaluation model to obtain a surface defect evaluation coefficient; the fabric texture defect grid is input into the texture defect evaluation model to obtain a texture defect evaluation coefficient; and the fabric color defect grid is input into the color defect evaluation model to obtain a color defect evaluation coefficient; and the fabric surface defect grid, the fabric texture defect grid, the fabric color defect grid, the surface defect evaluation coefficient, the texture defect evaluation coefficient and the color defect evaluation coefficient are collated to obtain the fabric quality evaluation atlas.

[0009] Optionally, surface feature convolution is performed according to the fabric production image to obtain a fabric surface feature grid; qualified fabric image sample retrieval is performed according to the fabric production requirement to obtain a qualified fabric image set; surface feature convolution is performed according to the qualified fabric image set to obtain a plurality of qualified surface feature grids; feature aggregation is performed on the plurality of qualified surface feature grids to obtain a surface feature reference grid; and surface defect detection is performed on the fabric surface feature grid according to the surface feature reference grid to obtain the fabric surface defect grid.

[0010] Optionally, surface defect path tracing is performed on the fabric quality evaluation graph according to the fabric brushing machine monitoring data to obtain a fabric surface defect path; texture defect path tracing is performed on the fabric quality evaluation graph according to the fabric brushing machine monitoring data to obtain a fabric texture defect path; color defect path tracing is performed on the fabric quality evaluation graph according to the fabric brushing machine monitoring data to obtain a fabric color defect path; and correlation feature analysis is performed on the fabric production control scheme according to the fabric surface defect path, the fabric texture defect path, and the fabric color defect path, respectively, to construct a fabric defect control correlation graph; and multi-parameter adjustment is performed on the fabric production control scheme according to the fabric defect control correlation graph to obtain the first fabric production adjustment domain.

[0011] Optionally, a fabric production adjustment first scheme is extracted according to the first fabric production adjustment domain; fabric feature prediction is performed based on the fabric production adjustment first scheme to obtain a first fabric feature prediction result; a first fabric quality evaluation sequence is obtained according to the fabric quality evaluation multi-dimensional model based on the first fabric feature prediction result; and if the first fabric quality evaluation sequence satisfies a fabric quality evaluation constraint, the fabric production adjustment first scheme is added to the first adjustment guide population.

[0012] Optionally, the fabric quality evaluation constraint includes a surface defect evaluation constraint, a texture defect evaluation constraint, and a color defect evaluation constraint.

[0013] Optionally, based on the fabric quality assessment multidimensional model, the first fabric production regulation domain is bred and optimized according to the first regulation-guided population to obtain a second regulation-guided population; based on the fabric quality assessment multidimensional model, the first fabric production regulation domain is bred and optimized according to the second regulation-guided population until multiple regulation-guided populations that meet a predetermined number of bred optimization attempts are obtained; the comprehensive fabric quality of the multiple regulation-guided populations is calculated according to the fabric quality assessment multidimensional weights to obtain the comprehensive fabric quality distribution; based on the comprehensive fabric quality distribution, the multiple regulation-guided populations are optimized for comprehensive quality according to the comprehensive fabric quality threshold to obtain a second fabric production regulation domain; and energy consumption minimization optimization is performed according to the second fabric production regulation domain to obtain the production control optimization result.

[0014] Optionally, the first regulation guide population is used to identify differences in the first fabric production regulation domain to obtain a first regulation difference vector set; random mutation is performed on the first regulation difference vector set to obtain a first regulation mutation vector set; mutation propagation is performed on the first fabric production regulation domain based on the first regulation mutation vector set to obtain a first regulation mutation propagation space; and fabric quality assessment optimization is performed on the first regulation mutation propagation space based on the fabric quality assessment multidimensional model to generate the second regulation guide population.

[0015] Secondly, this application also provides a machine vision-based fabric quality assessment system for napping machines, used to execute the machine vision-based fabric quality assessment method for napping machines as described in the first aspect. The machine vision-based fabric quality assessment system includes: a fabric production data acquisition module, used to simultaneously acquire fabric production images when the napping machine executes a fabric production instruction, the fabric production instruction including a fabric production control scheme corresponding to the fabric production requirements; and a multi-dimensional assessment module, used to introduce a multi-dimensional fabric quality assessment model, combine the fabric production requirements and the fabric production images to perform multi-dimensional fabric quality assessment, and construct a fabric quality assessment map. An adaptive adjustment module is used to adaptively adjust the fabric production control scheme according to the fabric quality assessment map to obtain a first fabric production adjustment domain; a quality assessment optimization module is used to optimize the fabric quality assessment of the first fabric production adjustment domain according to the fabric quality assessment multidimensional model to obtain a first adjustment guide population; a multiple reproduction optimization module is used to guide the first fabric production adjustment domain to perform multiple reproduction optimizations based on the fabric quality assessment multidimensional model and the first adjustment guide population to obtain production control optimization results; and an optimization control module is used to optimize the control of the napping machine according to the production control optimization results.

[0016] One or more technical solutions provided in this application have at least the following beneficial effects: When the napping machine executes a fabric production instruction, a fabric production image is simultaneously acquired, where the fabric production instruction includes a fabric production control scheme corresponding to the fabric production requirements; a multi-dimensional fabric quality assessment model is introduced, combining the fabric production requirements and the fabric production image to perform multi-dimensional fabric quality assessment and construct a fabric quality assessment map; the fabric production control scheme is adaptively adjusted according to the fabric quality assessment map to obtain a first fabric production adjustment domain; the first fabric production adjustment domain is optimized based on the multi-dimensional fabric quality assessment model to obtain a first adjustment guidance population; based on the multi-dimensional fabric quality assessment model, the first fabric production adjustment domain is guided by the first adjustment guidance population to undergo multiple breeding optimizations to obtain production control optimization results; and the napping machine is optimized and controlled based on the production control optimization results. In other words, by using a multi-dimensional fabric quality assessment model, comprehensively considering multiple dimensions of the fabric, a fabric quality assessment map is constructed, enabling adaptive adjustment of the napping machine's production control scheme. This allows for dynamic adjustment of the production process, avoiding instability during production, improving the stability of fabric quality, and thus increasing production efficiency.

[0017] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the method for evaluating the quality of napped fabrics using machine vision, as described in this application.

[0020] Figure 2 This is a schematic diagram of the structure of the napping machine fabric quality assessment system that incorporates machine vision, as described in this application.

[0021] Figure labeling: 11 Fabric production data acquisition module, 12 Multidimensional evaluation module, 13 Adaptive adjustment module, 14 Quality evaluation optimization module, 15 Multiple reproduction optimization module, 16 Optimization control module. Detailed Implementation

[0022] This application provides a method and system for evaluating the quality of napping machine fabrics using machine vision. This solves the technical problem in existing technologies where fabric quality evaluation relies on a single dimension of detection, lacking comprehensive evaluation across multiple dimensions, resulting in insufficient accuracy and consequently affecting fabric quality stability. By employing a multi-dimensional fabric quality evaluation model that comprehensively considers multiple dimensions of the fabric, a fabric quality evaluation map is constructed. This enables adaptive adjustment of the napping machine's production control scheme, allowing for dynamic adjustments to the production process, avoiding instability, improving fabric quality stability, and ultimately enhancing production efficiency.

[0023] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0024] Example 1, please refer to the appendix. Figure 1 This application provides a method for quality assessment of napped fabrics using machine vision, wherein the method is executed by a machine vision-based quality assessment system for napped fabrics, and specifically includes the following steps: When the napping machine executes the fabric production instruction, it simultaneously acquires the fabric production image. The fabric production instruction includes the fabric production control scheme corresponding to the fabric production requirements.

[0025] Furthermore, this application also includes the following steps: when the napping machine executes the fabric production instruction, it synchronously acquires fabric production monitoring images according to the machine vision device; and performs noise reduction processing on the fabric production monitoring images to obtain the fabric production images.

[0026] Specifically, when the production process starts, the napping machine executes corresponding operations according to preset fabric production instructions. These instructions include a production control scheme tailored to the fabric's production needs. In other words, when processing the fabric, the napping machine adjusts parameters such as rotation speed, pressure, and napping head tension according to the production instructions to achieve the target quality requirements. Simultaneously, a machine vision device installed on the napping machine acquires production monitoring images of the fabric. This machine vision device typically uses a high-resolution industrial camera to record the real-time state of the fabric during production by capturing images of the fabric surface, including information such as surface defects, napping effect, and pile distribution. Machine vision devices are based on camera and image processing technology and are typically used to automatically acquire and analyze image data during the production process. They can capture, process, and analyze images to help achieve functions such as automatic detection and quality assessment. Fabric production monitoring images refer to images acquired by the machine vision device, reflecting the actual state of the fabric at various moments during production, including information on surface quality, structure, and defects.

[0027] Because noise may exist during image acquisition, such as uneven lighting, lens defects, or camera noise, it is necessary to denoise the acquired fabric production monitoring images to remove unnecessary interference and ensure that useful information, such as fabric surface defects and fiber distribution, is clearly visible. For example, using Gaussian filtering or median filtering algorithms to denoise the images can remove 5% of noise points, improving the quality of the fabric production monitoring images. For instance, a machine vision device acquires fabric production images, each 1024*1024 pixels in size, with a resolution of 0.1 mm per pixel, acquiring images every 30 seconds. Using Gaussian filtering for denoising reduces the noise level to 3%, ensuring clear image edges for easier subsequent processing.

[0028] The fabric production instruction includes the fabric production control scheme corresponding to the fabric production requirements. The production control scheme sets the corresponding machine parameters according to the quality requirements of the target fabric to ensure that the fabric is always maintained within the predetermined quality standard range during the production process.

[0029] A multidimensional fabric quality assessment model is introduced, which combines the fabric production requirements and the fabric production images to conduct multidimensional assessment of fabric quality and construct a fabric quality assessment map.

[0030] Furthermore, this application also includes the following steps: performing multi-dimensional defect capture on the fabric production image according to the fabric production requirements to obtain a fabric surface defect grid, a fabric texture defect grid, and a fabric color defect grid; activating the fabric quality assessment multi-dimensional model, which includes a surface defect assessment model, a texture defect assessment model, and a color defect assessment model; inputting the fabric surface defect grid into the surface defect assessment model to obtain surface defect assessment coefficients; inputting the fabric texture defect grid into the texture defect assessment model to obtain texture defect assessment coefficients; inputting the fabric color defect grid into the color defect assessment model to obtain color defect assessment coefficients; and organizing the fabric surface defect grid, the fabric texture defect grid, the fabric color defect grid, the surface defect assessment coefficients, the texture defect assessment coefficients, and the color defect assessment coefficients to obtain the fabric quality assessment map.

[0031] Furthermore, this application also includes the following steps: performing surface feature convolution on the fabric production image to obtain a fabric surface feature grid; retrieving qualified fabric image samples according to the fabric production requirements to obtain a qualified fabric image set; performing surface feature convolution on the qualified fabric image set to obtain multiple qualified surface feature grids; performing feature aggregation on the multiple qualified surface feature grids to obtain a surface feature reference grid; and performing surface defect detection on the fabric surface feature grid based on the surface feature reference grid to obtain the fabric surface defect grid.

[0032] Specifically, surface feature convolution is performed on fabric production images. Using a sliding window approach, details are extracted from various parts of the image, capturing information such as fabric texture, pile distribution, and surface unevenness, forming a fabric surface feature grid that reflects the characteristic information of specific areas on the fabric surface. The convolution operation decomposes the image into multiple small regions, extracting local features such as texture, pile distribution, and surface smoothness through a sliding convolution kernel. These features are then synthesized into a fabric surface feature grid through a convolutional network, with each grid unit representing a portion of the fabric's surface features.

[0033] Based on fabric production requirements, qualified fabric image samples that meet the standards for fabric production are retrieved from existing image libraries to obtain a qualified fabric image set, showcasing the characteristics of fabrics that meet quality standards. The qualified fabric image set is a collection of images selected from existing fabric image libraries that meet production requirements, demonstrating the fabric surface that meets quality standards. The retrieved qualified fabric image set is then subjected to the same surface feature convolution to extract its surface features, generating multiple qualified surface feature grids that demonstrate the characteristic standards of ideal fabrics.

[0034] Multiple qualified surface feature grids are aggregated and merged into a surface feature reference grid, which integrates all standard fabric features in qualified images and serves as a standard reference for subsequent fabric quality inspection. Feature aggregation integrates useful feature information from multiple grids, such as texture patterns and pile density, removes redundancy and noise, and obtains a unified surface feature reference grid that represents the ideal features of the fabric surface.

[0035] Based on the generated surface feature reference mesh, the surface feature mesh of the fabric in actual production is compared to detect whether there are defects on the fabric surface. Through difference analysis, non-standard parts of the fabric surface are identified, such as uneven pile, scratches, color differences, etc., and a fabric surface defect mesh is generated to record the type and location of the defects on the fabric surface. For example, if the pile density of a certain part of the fabric is 6 piles / cm², which is lower than the standard of 8 piles / cm², a corresponding defect mesh is generated to indicate the uneven pile problem in this part.

[0036] In addition to surface defects, texture defects and color defects are also captured according to fabric production requirements, resulting in fabric texture defect meshes and fabric color defect meshes, respectively. The texture defect mesh captures the unevenness of the fabric surface texture, while the color defect mesh detects the inconsistency of the fabric color. For example, if a part of the fabric has a color difference, such as a color deviation exceeding the standard range, the color defect mesh will record this information and display the location and size of the color difference; similarly, if some parts of the fabric have uneven texture, the texture defect mesh will also record the corresponding information.

[0037] The multidimensional fabric quality assessment model is a machine learning model that comprehensively considers various aspects of fabric quality, including surface defect assessment models, texture defect assessment models, and color defect assessment models. Among them, the surface defect assessment model detects whether there are defects on the fabric surface, such as scratches, uneven pilling, and stains, by analyzing the surface features in the fabric image; the texture defect assessment model is used to identify whether the fabric texture is uniform and regular; and the color defect assessment model uses the color information of the image to detect whether there are color-related quality problems in the fabric, including color difference, uneven dyeing, and fading.

[0038] A surface defect assessment model is constructed by collecting a large amount of fabric surface image sample data. Each image is labeled with the defect type, such as scratches, color difference, uneven pilling, etc., as well as the location and area of ​​the defect. The image data is standardized to ensure uniform size and normalization, so that pixel values ​​range from 0 to 1. For example, suppose there are 2000 fabric surface images, of which 500 contain scratches, 500 contain uneven pilling, and the remaining 1000 are defect-free images. Each image is 256x256 pixels. All images are labeled, defining different defect categories and assigning corresponding labels to each image. For example, scratched images are labeled as 1, defect-free images as 0, and uneven pilling images as 2. The processed fabric surface image sample data is divided into training and validation sets, typically 80% for training and 20% for validation. A suitable convolutional neural network architecture is selected, such as a CNN model, including an input layer, convolutional layers, fully connected layers, and an output layer. The input layer receives preprocessed image data, i.e., the pixel values ​​of each image. Multiple convolutional layers are used to extract local features of the image; the kernel size is typically 3×3 or 5×5, and the stride is set to 1. Each convolutional layer is followed by a pooling layer, such as a max-pooling layer, to reduce the size of the feature map and decrease computation. The convolutional and pooled feature maps are flattened and fed into fully connected layers for high-level feature abstraction, ultimately outputting defect classification information. After flattening, the feature maps pass through two fully connected layers, each with 256 neurons. The output layer classifies surface defects according to their type. A softmax activation function is used to output the probability values ​​for each defect category.

[0039] The labeled training set is input into a CNN model, which is then trained using the backpropagation algorithm. During training, the model learns the features of different defects and adjusts the network parameters. For example, after 10 epochs of training, the model can predict the type of defect from an input image. After training, the model is evaluated using a validation set, and metrics such as accuracy, recall, and F1-score are calculated. Based on the evaluation results, model parameters are adjusted or different network architectures are tried to improve model performance. For example, the model achieves an accuracy of 85%, a recall of 80%, and an F1-score of 82%. If the results are unsatisfactory, more training data or a deeper network structure needs to be added for training. Through further optimization of the model, such as adjusting the number of convolutional layers and adding data augmentation, the test set accuracy is improved to 90%. The trained surface defect assessment model is deployed in a real-world production environment, receiving real-time images of fabric production captured by a machine vision device and automatically detecting surface defects in the images. The input to the surface defect assessment model is the real-time images captured during fabric production, i.e., a grid of surface defects on the fabric surface; based on the data in this grid, the model assesses the type and severity of defects and outputs surface defect assessment coefficients. The surface defect assessment coefficient is used to quantify the severity of surface defects in fabrics. It is typically a value between 0 and 1, with smaller values ​​indicating more severe defects and poorer quality. A coefficient higher than 0.8 may indicate a minor defect, while a coefficient lower than 0.2 indicates a more severe defect.

[0040] The training process for texture defect assessment models and color defect assessment models is similar. The texture defect assessment model takes a fabric texture defect mesh as input, focusing on the uniformity and regularity of the texture. The image resolution is typically 1024×1024, and the image may contain uneven texture distribution, texture distortion, etc. Similar to the surface defect model, the image needs to be denoised and standardized before input, and texture feature extraction may be performed to enhance the model's ability to perceive texture details. The output is a numerical value representing the severity of the fabric texture defect, typically between 0 and 1, with smaller values ​​indicating more severe defects. For example, 0.7 indicates a relatively small degree of texture defect, possibly with some uneven texture areas. The training process uses a regression loss function, such as mean squared error.

[0041] The color defect assessment model takes a fabric color defect grid as input, focusing on the fabric's color uniformity and color difference. The image resolution is typically 1024×1024, and the image may contain issues such as color difference and uneven dyeing. Color information needs to be converted to a color space, such as from RGB to HSV or Lab, so that the model can better understand and assess color difference. The output is a color defect assessment coefficient, quantifying the severity of the fabric color defect, usually a value between 0 and 1, with smaller values ​​indicating more severe color difference problems. For example, 0.25 indicates that the fabric has a significant color difference that will affect its appearance. If the specific location of the color defect is needed, the color difference area in the fabric will be marked. A CNN model is used, and the training process uses a regression loss function or a cross-entropy loss function.

[0042] A fabric surface defect mesh is input into the surface defect assessment model. This mesh records the location and type of surface defect areas, containing defect information for various regions of the fabric surface, such as scratches, stains, or uneven fuzzing. The surface defect assessment model learns the characteristics of the fabric surface defects and assesses the severity of the defects based on these characteristics. Through processing by the surface defect assessment model, a surface defect assessment coefficient is generated. This coefficient is a value between 0 and 1, reflecting the severity of the fabric surface defect. A coefficient higher than 0.8 indicates a minor defect, while a coefficient lower than 0.2 indicates a severe defect.

[0043] The fabric texture defect mesh is input into the texture defect evaluation model. This mesh records areas of uneven or distorted texture, including the distribution of the fabric surface texture, texture uniformity, and the presence of irregular textures such as distortion and breaks. The texture defect evaluation model analyzes the input mesh and identifies abnormal or defective areas in the fabric texture. By comparing the actual texture with the ideal texture, the model assesses the severity of the texture defect. The texture defect evaluation model outputs a texture defect evaluation coefficient, which quantifies the severity of the fabric texture defect; a smaller value indicates a more severe defect.

[0044] A fabric color defect mesh is input into the color defect assessment model. This mesh records color differences or uneven color distribution across different areas of the fabric surface, allowing for the inspection of color variations, uneven color distribution, or color variability. The color defect assessment model analyzes the input mesh to evaluate the presence of color differences, uneven color distribution, and other issues on the fabric surface, determining the severity of the color defects by learning the fabric's color patterns. The model outputs a color defect assessment coefficient, representing the degree of the fabric color defect. A smaller value indicates a more severe color difference problem.

[0045] By integrating the fabric surface defect grid, fabric texture defect grid, and fabric color defect grid with their corresponding evaluation coefficients—namely, surface defect evaluation coefficients, texture defect evaluation coefficients, and color defect evaluation coefficients—a comprehensive quality assessment model is formed. This ensures that defect information for each dimension is recorded in detail, resulting in a fabric quality assessment map. This map contains defect information for each dimension of the fabric surface, texture, and color, including the type, location, and evaluation coefficient for each defect. For example, a certain area might show: minor surface scratches (evaluation coefficient 0.9), uneven texture (evaluation coefficient 0.4), and color difference (evaluation coefficient 0.2).

[0046] Through multi-dimensional defect capture and quality assessment models, fabric quality is comprehensively evaluated from multiple dimensions, ensuring that any quality issues can be detected promptly. By generating fabric quality assessment maps, the fabric quality status can be fed back in real time, allowing for timely adjustments to production processes, preventing the generation of defective products, and improving the fabric's quality stability.

[0047] The fabric production control scheme is adaptively adjusted based on the fabric quality assessment map to obtain the first fabric production adjustment domain.

[0048] Furthermore, this application also includes the following steps: tracing the surface defect path of the fabric quality assessment map based on the napping machine monitoring data to obtain the fabric surface defect path; tracing the texture defect path of the fabric quality assessment map based on the napping machine monitoring data to obtain the fabric texture defect path; tracing the color defect path of the fabric quality assessment map based on the napping machine monitoring data to obtain the fabric color defect path; performing correlation feature analysis on the fabric production control scheme based on the fabric surface defect path, the fabric texture defect path, and the fabric color defect path respectively to construct a fabric defect control correlation map; and adjusting multiple parameters of the fabric production control scheme based on the fabric defect control correlation map to obtain the first fabric production adjustment domain.

[0049] Specifically, acquiring napping machine monitoring data involves collecting real-time data from the napping machine, including equipment operating parameters and fabric processing status, such as napping machine speed, pressure, abrasion degree of the abrasive, fabric tension, and production speed. These parameters reflect the fabric's processing status during production and thus affect fabric quality. The napping machine monitoring data is then used to trace surface defect paths in the fabric quality assessment map, tracing the formation process and possible causes of fabric surface defects. The purpose of defect path tracing is to find the specific causes of defects and provide a basis for subsequent production adjustments. For example, if a significant defect appears on the fabric surface in a certain area, such as scratches or stains, the correlation between the production data and the quality assessment results for that area is analyzed. Assuming a surface defect assessment coefficient of 0.4 for a specific area indicates significant surface scratches, comparing the napping machine data for that area reveals that the napping machine speed or pressure is too high, thus inferring the cause of the defect and finding the path leading to the surface defect. Through path tracing, the generated fabric surface defect path will include the specific factors and paths leading to the defect. For example, suppose that during a certain production process, the speed of the napping machine is set to 1500 rpm, the pressure is 120 N, and the fabric tension is 10 N. The fabric quality assessment chart shows that the surface defect assessment coefficient for a certain area is 0.4, indicating that there is a scratch problem in that area. The napping machine speed in this area is 1500 rpm, which is consistent with the known relationship that high speed may cause surface defects. Therefore, it is believed that high speed is the main cause of surface scratches, and it is recommended to reduce the speed to 1200 rpm to reduce the occurrence of similar defects.

[0050] Similarly, for texture defect path tracing and color defect path tracing, texture defect path tracing and color defect path tracing are performed on the fabric quality assessment map based on napping machine monitoring data to determine the specific causes of defects and obtain the fabric texture defect path and fabric color defect path. In other words, texture defect path tracing analysis analyzes the causes of uneven fabric texture or other texture defects, enabling the tracing of the source of texture defects. Analyzing the formation process of fabric color unevenness, color difference, and other problems, by combining napping machine monitoring data, the source of color defects and related production factors, such as dyeing processes and fabric tension, can be traced.

[0051] This paper establishes a fabric defect control correlation map by linking fabric surface defect paths, fabric texture defect paths, and fabric color defect paths with the fabric production control scheme. This map displays the relationship between different defects and production parameters. The fabric defect control correlation map connects fabric defect paths with the production control scheme for analysis and optimization of the production process. By analyzing the fabric defect control correlation map, it identifies which production parameters need adjustment to reduce specific types of defects. For example, if the fabric defect control correlation map shows that high rotation speed causes surface scratches, the rotation speed of the napping machine can be reduced, or the pressure of the napping machine can be adjusted to optimize the fabric surface quality. For texture and color defects, fabric quality can be improved by adjusting fabric tension, production speed, or abrasive wear conditions. By adjusting multiple parameters of the fabric production control scheme, a first fabric production adjustment domain is obtained, encompassing multiple fabric production adjustment schemes. Fabric quality is optimized by adjusting production parameters. For example, adjusting parameters such as the abrasive type, speed setting, and temperature control of the napping machine can improve fabric quality. By tracing the paths of fabric defects and constructing the fabric defect control correlation map, problems in the production process can be accurately identified, and precise parameter adjustments can be made.

[0052] Based on the multidimensional model for fabric quality assessment, the first fabric production regulation domain is optimized for fabric quality assessment to obtain the first regulation-guiding population.

[0053] Furthermore, this application also includes the following steps: extracting a first fabric production regulation scheme based on the first fabric production regulation domain; performing fabric feature prediction based on the first fabric production regulation scheme to obtain a first fabric feature prediction result; obtaining a first fabric quality assessment sequence based on the first fabric feature prediction result and the multidimensional model for fabric quality assessment; and adding the first fabric production regulation scheme to the first regulation guidance population if the first fabric quality assessment sequence satisfies the fabric quality assessment constraints.

[0054] Furthermore, this application also includes the following steps: the fabric quality assessment constraints include surface defect assessment constraints, texture defect assessment constraints, and color defect assessment constraints.

[0055] Specifically, a first fabric production adjustment scheme is arbitrarily extracted from the first fabric production adjustment domain, including specific production parameter settings such as napping machine speed, pressure, and temperature. Fabric feature prediction is performed based on the first fabric production adjustment scheme to obtain the first fabric feature prediction results, including surface defect feature prediction, texture defect feature prediction, and color defect feature prediction. The first fabric feature prediction results are evaluated using a multi-dimensional fabric quality assessment model to obtain a first fabric quality assessment sequence. That is, the surface defect feature prediction is input into the surface defect assessment model to obtain the surface defect assessment coefficient, the texture defect feature prediction is input into the surface defect assessment model to obtain the texture defect assessment coefficient, and the color defect feature prediction is input into the surface defect assessment model to obtain the color defect assessment coefficient. The surface defect assessment coefficient, texture defect assessment coefficient, and color defect assessment coefficient together constitute the first fabric quality assessment sequence. For example, the surface defect assessment coefficient is 0.9, the texture defect assessment coefficient is 0.6, and the color defect assessment coefficient is 0.8.

[0056] Fabric quality assessment constraints are a set of limiting conditions for fabric quality assessment results, including surface defect assessment constraints, texture defect assessment constraints, and color defect assessment constraints, used to ensure that the produced fabrics meet quality standards. Surface defect assessment constraints specify the maximum tolerance for surface defects, texture defect assessment constraints specify the maximum tolerance for texture defects, and color defect assessment constraints specify the maximum tolerance for color defects. For example, fabric quality assessment constraints may include: a surface defect assessment coefficient greater than 0.7, a texture defect assessment coefficient greater than 0.5, and a color defect assessment coefficient greater than 0.6.

[0057] Based on the fabric quality assessment constraints, check whether the first fabric quality assessment sequence meets the constraints. If it does, add the first fabric production regulation scheme corresponding to the first fabric quality assessment sequence to the first regulation guide population. For other fabric production regulation schemes in the first fabric production regulation domain, perform the aforementioned steps in the same way, and finally add all fabric production regulation schemes that meet the fabric quality assessment constraints to the first regulation guide population.

[0058] For example, assume that the first fabric production adjustment domain includes three adjustment schemes: adjustment scheme 1 has a rotation speed of 1200 rpm, a pressure of 100 N, and a fabric tension of 12 N; adjustment scheme 2 has a rotation speed of 1400 rpm, a pressure of 120 N, and a fabric tension of 14 N; and adjustment scheme 3 has a rotation speed of 1600 rpm, a pressure of 130 N, and a fabric tension of 15 N. Fabric feature prediction for adjustment scheme 1 shows almost no surface defects; texture prediction for adjustment scheme 1 shows moderate texture uniformity with some non-uniformity; and color prediction for adjustment scheme 1 shows good color consistency but slight color difference. Therefore, the first fabric feature prediction results include surface defect feature prediction, texture defect feature prediction, and color defect feature prediction. Inputting the first fabric feature prediction results into a multi-dimensional fabric quality assessment model, the evaluation coefficients for surface, texture, and color defects are calculated respectively, resulting in a surface defect evaluation coefficient of 0.9, a texture defect evaluation coefficient of 0.6, and a color defect evaluation coefficient of 0.8. Fabric quality assessment constraints: Surface defect assessment coefficient greater than 0.7, texture defect assessment coefficient greater than 0.5, color defect assessment coefficient greater than 0.6. The first fabric quality assessment sequence is compared with these constraints: the surface defect assessment coefficient of 0.9 is greater than 0.7, meeting the constraints; the texture defect assessment coefficient of 0.6 is greater than 0.5, meeting the constraints; and the color defect assessment coefficient of 0.8 is greater than 0.6, meeting the constraints. Since the first fabric quality assessment sequence satisfies all quality assessment constraints, this scheme meets the quality requirements. The first fabric production adjustment scheme (speed 1200 rpm, pressure 100 N, fabric tension 12 N) is added to the first adjustment guide population. Repeating the aforementioned steps, adjustment scheme 2 has a surface defect assessment coefficient of 0.8, a texture defect assessment coefficient of 0.7, and a color defect assessment coefficient of 0.9; adjustment scheme 3 has a surface defect assessment coefficient of 0.6, a texture defect assessment coefficient of 0.4, and a color defect assessment coefficient of 0.7. The quality assessment sequence of adjustment scheme 2 satisfies the fabric quality assessment constraints, and adjustment scheme 2 is added to the first adjustment guide population. Since the quality assessment sequence of regulation scheme 3 does not meet the surface defect assessment constraints and texture defect assessment constraints, regulation scheme 3 is not added to the first regulation guide population.

[0059] By predicting adjustment schemes based on quality assessment, the quality characteristics of each fabric can be precisely controlled, ensuring that optimal quality standards are met at every stage of the production process. This ensures that each adjustment scheme meets quality assessment constraints, minimizing fabric defects and improving the stability and consistency of fabric quality.

[0060] Based on the multidimensional model for fabric quality assessment, the first fabric production regulation domain is guided by the first regulatory guiding population to undergo multiple breeding optimizations to obtain production control optimization results.

[0061] Furthermore, this application also includes the following steps: based on the fabric quality assessment multidimensional model, performing reproductive optimization on the first fabric production regulation domain according to the first regulation guiding population to obtain a second regulation guiding population; based on the fabric quality assessment multidimensional model, continuing to perform reproductive optimization on the first fabric production regulation domain according to the second regulation guiding population until multiple regulation guiding populations that meet a predetermined number of reproductive optimization attempts are obtained; calculating the comprehensive fabric quality of the multiple regulation guiding populations according to the fabric quality assessment multidimensional weights to obtain the comprehensive fabric quality distribution; based on the comprehensive fabric quality distribution, performing comprehensive quality optimization on the multiple regulation guiding populations according to the comprehensive fabric quality threshold to obtain a second fabric production regulation domain; and performing energy consumption minimization optimization on the second fabric production regulation domain to obtain the production control optimization result.

[0062] Furthermore, this application also includes the following steps: identifying differences in the first fabric production regulation domain based on the first regulation-guided population to obtain a first regulation difference vector set; performing random mutation on the first regulation difference vector set to obtain a first regulation mutation vector set; performing mutation propagation on the first fabric production regulation domain based on the first regulation mutation vector set to obtain a first regulation mutation propagation space; and performing fabric quality assessment optimization on the first regulation mutation propagation space based on the fabric quality assessment multidimensional model to generate a second regulation-guided population.

[0063] Specifically, the first regulation-guided population contains multiple regulation schemes that meet fabric quality assessment constraints. The first fabric production regulation domain includes multiple regulation schemes obtained after multi-parameter adjustments to the fabric production control schemes. By comparing the differences between each regulation scheme in the first regulation-guided population and the regulation schemes in the first fabric production regulation domain, significantly different regulation schemes are identified, which could be differences in parameters or differences in quality assessment results. The first regulation difference vector set is a set of vectors obtained through difference identification; each vector represents the difference between a regulation scheme and other schemes, representing information on the diversity of regulation schemes.

[0064] Random mutation is performed on the first set of adjustment difference vectors to change the parameters of some adjustment schemes, such as randomly adjusting the values ​​of production parameters like rotation speed, pressure, and fabric tension, generating new adjustment vectors and creating the first set of adjustment mutation vectors. Random mutation generates new schemes by randomly adjusting certain parameters of the adjustment schemes, exploring new production adjustment schemes to improve fabric quality. The first fabric production adjustment domain is then subjected to mutation propagation based on the first set of adjustment mutation vectors. Through propagation operations, adjustment schemes in the first fabric production adjustment domain are combined and expanded to generate multiple possible new schemes. New adjustment schemes are formed by combining different mutation vectors from the first set of adjustment mutation vectors. For example, the parameters of adjustment scheme A and adjustment scheme B can be combined through propagation operations to generate a new adjustment scheme C. For instance, adjustment scheme A has a rotation speed of 1500 rpm, pressure of 100 N, and tension of 12 N, while adjustment scheme B has a rotation speed of 1300 rpm, pressure of 110 N, and tension of 14 N. After mutation propagation, the new adjustment scheme C has a rotation speed of 1400 rpm, pressure of 105 N, and tension of 13 N. The first regulatory mutation and reproduction space contains all possible regulatory schemes, representing new combinations of production parameters generated from the mutation and reproduction process. Schemes in the first regulatory mutation and reproduction space are obtained by mutating the schemes in the first regulatory mutation vector set in a better direction.

[0065] The fabric quality assessment multidimensional model is used to optimize the first regulatory variation propagation space. This involves predicting fabric characteristics for each scheme in the first regulatory variation propagation space, obtaining corresponding surface defect feature predictions, texture defect feature predictions, and color defect feature predictions. The multidimensional model is then used to evaluate these predictions, obtaining the corresponding surface defect evaluation coefficients, texture defect evaluation coefficients, and color defect evaluation coefficients for each scheme. It is then determined whether these constraints are met. If they are, the scheme is added to the second regulatory guide population. The second regulatory guide population is a new population generated after quality assessment optimization, containing better regulatory schemes. Through difference identification, random mutation, and propagation operations, diverse regulatory schemes are generated. The quality of these new schemes is evaluated, and those that meet the fabric quality assessment constraints are selected, ensuring that each regulatory scheme meets quality requirements, thereby improving the consistency and quality stability of the final product.

[0066] The second guiding population continues to breed and optimize within the first fabric production regulation domain. The bred regulation schemes are then evaluated using a multi-dimensional fabric quality assessment model. This process is repeated until a predetermined number of breeding optimization attempts is reached. The predetermined number of breeding optimization attempts is a pre-set number of breeding attempts during the optimization process, typically to ensure the generation of a sufficient number of regulation schemes. Through multiple breeding optimization attempts, the most suitable regulation scheme is ultimately found. Multiple breeding and optimization attempts are performed according to the predetermined number of attempts. After each breeding attempt, a new regulation scheme is generated and its quality is evaluated until the predetermined number of breeding attempts is reached, such as 10. Each breeding attempt generates a new regulation scheme, which is then screened and optimized using a multi-dimensional fabric quality assessment model, gradually selecting the best regulation scheme, ultimately forming multiple guiding populations. These multiple guiding populations represent the set of optimal regulation schemes based on quality assessment during the production regulation process.

[0067] In fabric quality assessment, each dimension has a different weight influencing the final quality, resulting in a multi-dimensional weighting for fabric quality assessment. A higher weight value indicates a greater contribution of that dimension to the overall fabric quality. Specifically, the multi-dimensional weighting for fabric quality assessment includes weights for surface defect assessment, texture defect assessment, and color defect assessment. The overall fabric quality is calculated for multiple moderating guide populations, resulting in multiple overall fabric quality coefficients that constitute the overall fabric quality distribution. The overall fabric quality coefficient for each moderating scheme is calculated using the following formula: Overall Fabric Quality Coefficient = Surface Defect Assessment Coefficient × Surface Defect Assessment Weight + Texture Defect Assessment Coefficient × Texture Defect Assessment Weight + Color Defect Coefficient × Color Defect Assessment Weight. The overall fabric quality coefficient for each moderating scheme reflects its overall performance across all quality dimensions.

[0068] The overall fabric quality threshold is a set quality standard value representing the quality requirements of the fabric in various dimensions. Only adjustment schemes exceeding the overall fabric quality threshold are considered to meet the quality requirements. Multiple overall fabric quality coefficients corresponding to multiple adjustment guide populations in the overall fabric quality distribution are compared with the overall fabric quality threshold. If a coefficient is greater than or equal to the overall fabric quality threshold, the scheme meets the quality requirements and is added to the second fabric production adjustment domain. Energy consumption minimization optimization is performed on the second fabric production adjustment domain, evaluating the energy consumption of each scheme and finding the adjustment scheme with the minimum energy consumption to obtain the production control optimization result. For example, in one instance, the preset weights for surface defect evaluation, texture defect evaluation, and color defect evaluation are 0.4, 0.3, and 0.3, respectively. The surface defect evaluation coefficient for guide population 1 is adjusted to 0.9, the texture defect evaluation coefficient to 0.6, and the color defect evaluation coefficient to 0.7, resulting in a comprehensive quality coefficient of 0.75. Guide population 2 is adjusted to a rotation speed of 1500 rpm, a pressure of 120 N, and a fabric tension of 14 N, with surface defect evaluation coefficients of 0.7, texture defect evaluation coefficients of 0.6, and color defect evaluation coefficients of 0.5, resulting in a comprehensive quality coefficient of 0.61. Guide population 3 is adjusted to a rotation speed of 1700 rpm, a pressure of 140 N, and a fabric tension of 15 N, with surface defect evaluation coefficients of 0.5, texture defect evaluation coefficients of 0.7, and color defect evaluation coefficients of 0.4, resulting in a comprehensive quality coefficient of 0.53. Assuming the comprehensive fabric quality threshold is 0.6, both guide populations 1 and 2 meet the quality requirements and enter the second fabric production adjustment domain. Assuming the energy consumption of the guiding population 1 is 80 kWh and the energy consumption of the guiding population 2 is 90 kWh, then guiding population 1 is selected as the production control optimization result. Through comprehensive quality assessment and energy consumption optimization, it is ensured that the final selected regulation scheme not only meets the quality requirements but also minimizes energy consumption and improves production efficiency.

[0069] The grinding machine is optimized and controlled based on the production control optimization results.

[0070] Specifically, based on the production control optimization results, the parameters of the napping machine are adjusted, including the machine's rotation speed, pressure, abrasion degree of the abrasive, fabric tension, and production speed. The napping machine produces fabric according to the modified operating parameters, while the fabric production process is monitored in real time. Parameters such as fabric surface quality and texture uniformity are adjusted based on real-time feedback data to ensure that the fabric quality consistently meets requirements within each production cycle. If real-time feedback indicates that production parameters need adjustment, such as due to equipment wear or environmental changes, the napping machine's operating parameters are dynamically adjusted to ensure continuous optimization of quality and energy efficiency. Based on production monitoring data, adaptive adjustments are made according to the actual operating status of the napping machine, such as adjusting the rotation speed to adapt to different fabric characteristics or production conditions. Through optimized control of the napping machine, the surface quality, texture, and color of the fabric are ensured to meet predetermined standards. The production process no longer relies on manual experience but is dynamically adjusted based on data, thereby improving the consistency and stability of fabric quality. While ensuring quality, energy consumption is minimized through optimization and control, reducing energy consumption during the production process and lowering production costs.

[0071] In summary, the machine vision-based fabric quality assessment method for napping machines provided in this application has the following beneficial effects: When the napping machine executes a fabric production instruction, a fabric production image is simultaneously acquired, where the fabric production instruction includes a fabric production control scheme corresponding to the fabric production requirements; a multi-dimensional fabric quality assessment model is introduced, combining the fabric production requirements and the fabric production image to perform multi-dimensional fabric quality assessment and construct a fabric quality assessment map; the fabric production control scheme is adaptively adjusted based on the fabric quality assessment map to obtain a first fabric production adjustment domain; the first fabric production adjustment domain is optimized based on the multi-dimensional fabric quality assessment model to obtain a first adjustment guidance population; based on the multi-dimensional fabric quality assessment model, the first fabric production adjustment domain is guided by the first adjustment guidance population to undergo multiple breeding optimizations to obtain production control optimization results; and the napping machine is optimized and controlled based on the production control optimization results. In other words, by using a multi-dimensional fabric quality assessment model, which comprehensively considers multiple dimensions of the fabric, a fabric quality assessment map is constructed, enabling adaptive adjustment of the napping machine production control scheme. This allows the production process to be dynamically adjusted, avoiding instability in the production process, improving the stability of fabric quality, and thus increasing production efficiency.

[0072] Example 2: Based on the same inventive concept as the machine vision-integrated fabric quality assessment method for napped fabrics in Example 1, this application also provides a machine vision-integrated fabric quality assessment system for napped fabrics. Please refer to the appendix. Figure 2 The machine vision-integrated fabric quality assessment system for napping machines includes: The fabric production data acquisition module 11 is used to synchronously acquire fabric production images when the napping machine executes fabric production instructions, wherein the fabric production instructions include fabric production control schemes corresponding to fabric production requirements; the multi-dimensional evaluation module 12 is used to introduce a multi-dimensional fabric quality evaluation model, combine the fabric production requirements and the fabric production images to perform multi-dimensional fabric quality evaluation, and construct a fabric quality evaluation map; the adaptive adjustment module 13 is used to adaptively adjust the fabric production control scheme according to the fabric quality evaluation map to obtain a first fabric production adjustment domain; the quality evaluation optimization module 14 is used to perform fabric quality evaluation optimization on the first fabric production adjustment domain according to the multi-dimensional fabric quality evaluation model to obtain a first adjustment guide population; the multiple reproduction optimization module 15 is used to guide the first fabric production adjustment domain to perform multiple reproduction optimizations based on the multi-dimensional fabric quality evaluation model and the first adjustment guide population to obtain production control optimization results; and the optimization control module 16 is used to optimize the control of the napping machine according to the production control optimization results.

[0073] Furthermore, the fabric production data acquisition module 11 in the fabric quality assessment system for napping machines combined with machine vision is also used to: when the napping machine executes the fabric production instruction, synchronously acquire fabric production monitoring images according to the machine vision device; and perform noise reduction processing on the fabric production monitoring images to obtain the fabric production images.

[0074] Furthermore, the multi-dimensional evaluation module 12 in the machine vision-integrated napping machine fabric quality evaluation system is also used for: capturing multi-dimensional defects in the fabric production image according to the fabric production requirements to obtain fabric surface defect grids, fabric texture defect grids, and fabric color defect grids; activating the fabric quality evaluation multi-dimensional model, which includes a surface defect evaluation model, a texture defect evaluation model, and a color defect evaluation model; inputting the fabric surface defect grids into the surface defect evaluation model to obtain surface defect evaluation coefficients; inputting the fabric texture defect grids into the texture defect evaluation model to obtain texture defect evaluation coefficients; inputting the fabric color defect grids into the color defect evaluation model to obtain color defect evaluation coefficients; and organizing the fabric surface defect grids, fabric texture defect grids, fabric color defect grids, surface defect evaluation coefficients, texture defect evaluation coefficients, and color defect evaluation coefficients to obtain the fabric quality evaluation map.

[0075] Furthermore, the multi-dimensional evaluation module 12 in the machine vision-integrated napping machine fabric quality evaluation system is also used for: performing surface feature convolution on the fabric production image to obtain a fabric surface feature grid; retrieving qualified fabric image samples according to the fabric production requirements to obtain a qualified fabric image set; performing surface feature convolution on the qualified fabric image set to obtain multiple qualified surface feature grids; performing feature aggregation on the multiple qualified surface feature grids to obtain a surface feature reference grid; and performing surface defect detection on the fabric surface feature grid based on the surface feature reference grid to obtain the fabric surface defect grid.

[0076] Furthermore, the adaptive adjustment module 13 in the machine vision-integrated napping machine fabric quality assessment system is also used for: tracing surface defect paths in the fabric quality assessment map based on napping machine monitoring data to obtain fabric surface defect paths; tracing texture defect paths in the fabric quality assessment map based on napping machine monitoring data to obtain fabric texture defect paths; tracing color defect paths in the fabric quality assessment map based on napping machine monitoring data to obtain fabric color defect paths; performing correlation feature analysis on the fabric production control scheme based on the fabric surface defect paths, fabric texture defect paths, and fabric color defect paths to construct a fabric defect control correlation map; and adjusting multiple parameters of the fabric production control scheme based on the fabric defect control correlation map to obtain the first fabric production adjustment domain.

[0077] Furthermore, the quality assessment optimization module 14 in the machine vision-integrated napping machine fabric quality assessment system is also used for: extracting a first fabric production adjustment scheme based on the first fabric production adjustment domain; performing fabric feature prediction based on the first fabric production adjustment scheme to obtain a first fabric feature prediction result; obtaining a first fabric quality assessment sequence based on the first fabric feature prediction result and the fabric quality assessment multidimensional model; and adding the first fabric production adjustment scheme to the first adjustment guidance population if the first fabric quality assessment sequence satisfies the fabric quality assessment constraints.

[0078] Furthermore, the quality assessment optimization module 14 in the machine vision-integrated napping machine fabric quality assessment system is also used for: the fabric quality assessment constraints include surface defect assessment constraints, texture defect assessment constraints, and color defect assessment constraints.

[0079] Furthermore, the multiple reproduction optimization module 15 in the machine vision-integrated napping machine fabric quality assessment system is also used for: based on the fabric quality assessment multidimensional model, performing reproduction optimization on the first fabric production regulation domain according to the first regulation guiding population to obtain a second regulation guiding population; based on the fabric quality assessment multidimensional model, continuing to perform reproduction optimization on the first fabric production regulation domain according to the second regulation guiding population until multiple regulation guiding populations that meet the predetermined reproduction optimization number are obtained; calculating the comprehensive fabric quality of the multiple regulation guiding populations according to the fabric quality assessment multidimensional weights to obtain the comprehensive fabric quality distribution; based on the comprehensive fabric quality distribution, performing comprehensive quality optimization on the multiple regulation guiding populations according to the comprehensive fabric quality threshold to obtain a second fabric production regulation domain; and performing energy consumption minimization optimization based on the second fabric production regulation domain to obtain the production control optimization result.

[0080] Furthermore, the multiple reproduction optimization module 15 in the machine vision-integrated napping machine fabric quality assessment system is also used for: identifying differences in the first fabric production regulation domain based on the first regulation guide population to obtain a first regulation difference vector set; performing random mutation based on the first regulation difference vector set to obtain a first regulation mutation vector set; performing mutation reproduction on the first fabric production regulation domain based on the first regulation mutation vector set to obtain a first regulation mutation reproduction space; and performing fabric quality assessment optimization on the first regulation mutation reproduction space based on the fabric quality assessment multidimensional model to generate a second regulation guide population.

[0081] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Figure 1 The machine vision-integrated brushed fabric quality assessment method and specific examples in Example 1 are also applicable to the machine vision-integrated brushed fabric quality assessment system of this embodiment. Through the foregoing detailed description of the machine vision-integrated brushed fabric quality assessment method, those skilled in the art can clearly understand the machine vision-integrated brushed fabric quality assessment system of this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0082] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0083] Obviously, those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A method for evaluating the quality of napped fabrics using machine vision, characterized in that, include: When the napping machine executes the fabric production instruction, it simultaneously acquires the fabric production image. The fabric production instruction includes the fabric production control scheme corresponding to the fabric production requirements. A multidimensional fabric quality assessment model is introduced, and the fabric quality is assessed in a multidimensional way by combining the fabric production requirements and the fabric production images, thus constructing a fabric quality assessment map. The fabric production control scheme is adaptively adjusted based on the fabric quality assessment map to obtain the first fabric production adjustment domain. Based on the multidimensional fabric quality assessment model, the first fabric production regulation domain is optimized for fabric quality assessment to obtain the first regulation guide population. Based on the multidimensional model for fabric quality assessment, the first fabric production regulation domain is guided by the first regulation-guiding population to perform multiple breeding optimizations to obtain production control optimization results. The grinding machine is optimized and controlled based on the production control optimization results.

2. The method for quality assessment of napped fabrics combined with machine vision as described in claim 1, characterized in that, A multidimensional fabric quality assessment model is introduced, combining the fabric production requirements and the fabric production images to perform multidimensional fabric quality assessment, and constructing a fabric quality assessment map, including: Based on the fabric production requirements, multidimensional defect capture is performed on the fabric production image to obtain fabric surface defect mesh, fabric texture defect mesh, and fabric color defect mesh. Activate the multidimensional fabric quality assessment model, which includes a surface defect assessment model, a texture defect assessment model, and a color defect assessment model; The surface defect mesh of the fabric is input into the surface defect evaluation model to obtain the surface defect evaluation coefficients; Input the fabric texture defect mesh into the texture defect evaluation model to obtain texture defect evaluation coefficients; The fabric color defect grid is input into the color defect evaluation model to obtain the color defect evaluation coefficient; By organizing the fabric surface defect grid, the fabric texture defect grid, the fabric color defect grid, the surface defect evaluation coefficient, the texture defect evaluation coefficient, and the color defect evaluation coefficient, the fabric quality evaluation map is obtained.

3. The method for quality assessment of napped fabrics combined with machine vision as described in claim 2, characterized in that, Based on the fabric production requirements, multi-dimensional defect capture is performed on the fabric production image, including: Surface feature convolution is performed on the fabric production image to obtain a fabric surface feature mesh; Based on the fabric production requirements, a qualified fabric image sample retrieval is performed to obtain a qualified fabric image set. Perform surface feature convolution on the qualified fabric image set to obtain multiple qualified surface feature grids; The multiple qualified surface feature meshes are aggregated to obtain a surface feature reference mesh; Surface defect detection is performed on the fabric surface feature grid based on the surface feature reference grid to obtain the fabric surface defect grid.

4. The method for quality assessment of napped fabrics combined with machine vision as described in claim 1, characterized in that, The fabric production control scheme is adaptively adjusted based on the fabric quality assessment map to obtain a first fabric production adjustment domain, including: Based on the monitoring data of the napping machine, the surface defect path of the fabric quality assessment map is traced to obtain the surface defect path of the fabric. Based on the monitoring data of the napping machine, the texture defect path of the fabric quality assessment map is traced to obtain the fabric texture defect path; Based on the monitoring data of the napping machine, the color defect path of the fabric quality assessment map is traced to obtain the color defect path of the fabric. Based on the fabric surface defect path, the fabric texture defect path, and the fabric color defect path, the fabric production control scheme is analyzed for correlation features to construct a fabric defect control correlation map; The fabric production control scheme is adjusted by multiple parameters based on the fabric defect control correlation map to obtain the first fabric production adjustment domain.

5. The method for quality assessment of napped fabrics combined with machine vision as described in claim 1, characterized in that, Based on the aforementioned multidimensional fabric quality assessment model, the first fabric production regulation domain is optimized for fabric quality assessment to obtain a first regulation-guiding population, including: Based on the first fabric production adjustment domain, extract the first fabric production adjustment scheme; Based on the first scheme for adjusting fabric production, fabric characteristics are predicted to obtain the first fabric characteristic prediction result. Based on the first fabric feature prediction result, and according to the fabric quality assessment multidimensional model, a first fabric quality assessment sequence is obtained; If the first fabric quality assessment sequence satisfies the fabric quality assessment constraints, the first fabric production regulation scheme is added to the first regulation guide population.

6. The method for quality assessment of napped fabrics combined with machine vision as described in claim 1, characterized in that, Based on the aforementioned multidimensional model for fabric quality assessment, the first fabric production regulation domain is guided by the first regulatory guiding population to undergo multiple breeding optimizations to obtain production control optimization results, including: Based on the multidimensional model for fabric quality assessment, a second regulatory guide population is obtained by breeding and optimizing the first fabric production regulation domain according to the first regulatory guide population. Based on the multidimensional model for fabric quality assessment, the first fabric production regulation domain is further bred and optimized according to the second regulatory guide population until multiple regulatory guide populations that meet the predetermined number of bred optimizations are obtained. The overall fabric quality distribution is obtained by calculating the overall fabric quality of the multiple regulatory guide populations based on the multidimensional weights of fabric quality assessment. Based on the overall fabric quality distribution, the overall quality of the multiple regulatory guide populations is optimized according to the overall fabric quality threshold to obtain the second fabric production regulation domain. The energy consumption minimization optimization is performed based on the second fabric production adjustment domain to obtain the production control optimization result.

7. The method for quality assessment of napped fabrics combined with machine vision as described in claim 6, characterized in that, Based on the aforementioned multidimensional model for fabric quality assessment, a second regulatory guide population is obtained by performing reproductive optimization on the first fabric production regulation domain according to the first regulatory guide population, including: Based on the first regulatory guidance population, the first fabric production regulation domain is differentiated to obtain the first regulation difference vector set; Random mutation is performed on the first set of adjustment difference vectors to obtain the first set of adjustment mutation vectors; Based on the first set of regulation mutation vectors, the first fabric production regulation domain is mutated and multiplied to obtain the first regulation mutation multiplication space; Based on the fabric quality assessment multidimensional model, the fabric quality assessment optimization is performed on the first regulatory variation reproduction space to generate the second regulatory guide population.

8. The method for quality assessment of napped fabrics combined with machine vision as described in claim 5, characterized in that, The fabric quality assessment constraints include surface defect assessment constraints, texture defect assessment constraints, and color defect assessment constraints.

9. The method for quality assessment of napped fabrics combined with machine vision as described in claim 1, characterized in that, When the napping machine executes the fabric production instruction, it simultaneously acquires fabric production images, including: When the napping machine executes the fabric production instruction, it simultaneously acquires fabric production monitoring images according to the machine vision device. The fabric production monitoring image is denoised to obtain the fabric production image.

10. A quality assessment system for napped fabrics combined with machine vision, characterized in that, The step of implementing the machine vision-integrated brushed fabric quality assessment method according to any one of claims 1 to 9, wherein the machine vision-integrated brushed fabric quality assessment system comprises: The fabric production data acquisition module is used to synchronously acquire fabric production images when the napping machine executes a fabric production instruction, wherein the fabric production instruction includes a fabric production control scheme corresponding to the fabric production requirements. The multidimensional assessment module is used to introduce a multidimensional fabric quality assessment model, combine the fabric production requirements and the fabric production images to conduct multidimensional fabric quality assessment, and construct a fabric quality assessment map. An adaptive adjustment module is used to adaptively adjust the fabric production control scheme according to the fabric quality assessment map to obtain a first fabric production adjustment domain. The quality assessment optimization module is used to perform fabric quality assessment optimization on the first fabric production regulation domain according to the fabric quality assessment multidimensional model to obtain the first regulation guiding population. The multiple breeding optimization module is used to perform multiple breeding optimizations based on the fabric quality assessment multidimensional model and the first fabric production regulation domain guided by the first regulation guiding population to obtain production control optimization results. An optimization control module is used to optimize the control of the grinding machine based on the production control optimization results.

Images