Color firmness evaluation method and system

Abstract

The invention relates to the technical field of data analysis and evaluation, in particular to a color fastness evaluation method and system, and the method comprises the following steps: setting test conditions to simulate external factors encountered in the actual use or processing process of textiles; under the simulated test condition, treating the textile sample for a preset time; acquiring a multispectral image of the textile by adopting a multispectral imaging technology, and capturing color change of the textile; developing a self-adaptive image enhancement algorithm, and performing preprocessing; comparing spectral information of the multi-spectral image of the textile before and after processing based on chromatic aberration evaluation analysis so as to quantify the color fastness; analyzing the multispectral image after image enhancement by using a deep learning technology, extracting color attenuation features, and comprehensively evaluating a color fastness score; according to the color fastness evaluation method, basic color difference information is considered, advanced features such as textures, edges and local color modes of the textile are further included, and the accuracy and reliability of color fastness evaluation are remarkably improved.

Description

[0001] This application is a divisional application of the application filed on July 4, 2024, with application number 202410892426.6 and invention name “A method and system for evaluating color fastness”. Technical Field

[0002] The present invention relates to the technical field of data analysis and evaluation, and in particular to a color fastness evaluation method and system. Background Art

[0003] In the textile industry, color fastness is one of the important indicators for measuring the quality of textiles. It reflects the stability of the color of textiles when they are subjected to external factors such as washing, friction, and light. Traditional color fastness testing methods usually involve a series of standardized laboratory tests, such as friction tests, washing tests, and exposure tests. Although these methods can provide certain color fastness information, they have some limitations: first, they are often time-consuming and cannot achieve rapid evaluation; second, traditional testing methods often focus on a specific external influencing factor, which makes it difficult to fully reflect the color fastness performance of textiles in complex use environments; finally, these methods often require destructive sampling and cannot achieve non-destructive testing of textiles.

[0004] With the advancement of science and technology, the development of multispectral imaging technology and deep learning algorithms has provided new possibilities for the evaluation of textile color fastness. Multispectral imaging technology can capture images of textiles at different wavelengths, providing richer color and texture information than traditional RGB images, which helps to more accurately analyze the color characteristics of textiles. Deep learning algorithms, especially generative adversarial networks (GANs) and convolutional neural networks (CNNs), can effectively process and analyze these large amounts of multispectral image data and automatically extract key features that affect color fastness.

[0005] Therefore, a new method for evaluating color fastness of textiles can be constructed by combining multispectral imaging technology, adaptive image enhancement algorithm, and deep learning analysis method. Summary of the invention

[0006] Based on the above purpose, the present invention provides a color fastness evaluation method and system.

[0007] A method for evaluating color fastness comprises the following steps:

[0008] S1: Set test conditions to simulate external factors encountered by textiles during actual use or processing, including setting corresponding conditions of extrusion, friction, washing, exposure and sweat stains;

[0009] S2: The textile samples are treated for a predetermined time under simulated test conditions to simulate actual wear conditions;

[0010] S3: Use multispectral imaging technology to obtain multispectral images of textiles and capture the color changes of textiles;

[0011] S4: Develop an adaptive image enhancement algorithm to preprocess the multispectral images from S3 to cope with the non-uniform lighting and shadow issues of textile images;

[0012] S5: Based on color difference evaluation analysis, the spectral information of multispectral images of textiles before and after treatment is compared to quantify color fastness, considering the effects of fiber type, dye properties and test conditions on color fastness;

[0013] S6: Apply deep learning technology to analyze the multispectral image after image enhancement, extract color attenuation features, and combine with the quantitative color fastness data in S5 to comprehensively evaluate the color fastness score;

[0014] S7: Develop visualization tools to present color fastness evaluation results in a graphical manner.

[0015] Furthermore, the S1 specifically includes:

[0016] S11, setting of compression conditions: simulating the pressure on the textile during use, using standardized weights to compress the textile sample for a predetermined time, and the mass of the weights and the compression time are selected according to the actual use environment of the textile;

[0017] S12, setting of friction conditions: simulating the wear of textiles in daily use, repeatedly rubbing the surface of textile samples through a dedicated wear test instrument. The number and intensity of friction are set according to the standard test method to simulate the degree of wear in actual wearing or use;

[0018] S13, setting of washing conditions: simulating the washing fastness of textiles during the washing process by washing the textile samples multiple times in a standardized washing liquid. The number of washes, water temperature and type of detergent used are determined according to the material of the textiles and the expected use environment;

[0019] S14, setting of exposure conditions: simulating the light fastness of textiles under sunlight, by using ultraviolet lamps or exposing textile samples to natural light for a predetermined period to evaluate the effect of long-term or short-term exposure on the color of textiles. The exposure time and light intensity are adjusted according to the use scenario and light fastness requirements of the textiles;

[0020] S15, setting of sweat conditions: simulating the contact between textiles and human sweat, by treating the textile samples with simulated sweat samples.

[0021] Furthermore, the S3 specifically includes:

[0022] S31: Select a multispectral imaging system equipped with multiple filters of different wavelengths to ensure coverage of a wide band range from ultraviolet to near infrared, and capture textile images at multiple wavelengths to obtain the response of textiles under different spectra;

[0023] S32: Before imaging, ensure that the textile sample is flat and under standardized lighting conditions, and calibrate the imaging system using a standard color card to ensure the color accuracy of the acquired image;

[0024] S33: Imaging the textile sample at multiple predetermined wavelengths, collecting one image at each wavelength, and sequentially capturing images at each wavelength by changing the filters in the imaging system to construct a complete multispectral data set;

[0025] S34: De-noising, contrast adjustment and color correction are performed on the collected multi-wavelength images, and the processed images are synthesized into a complete multi-spectral image, which includes the response information of the textile in the entire spectral range.

[0026] S35: Identify color changes of textiles at different wavelengths by comparing multispectral images before and after processing.

[0027] Furthermore, the adaptive image enhancement algorithm in S4 specifically includes:

[0028] Adaptive local brightness adjustment: An improved local brightness adjustment algorithm is used to automatically adjust the brightness and contrast according to the specific area of ​​the textile image, taking into account the texture and fiber structure of the textile, ensuring that the brightness adjustment process does not destroy the details of the textile or introduce unnatural visual effects;

[0029] Intelligent shadow compensation mechanism: Identify shadow areas in multispectral images and perform intelligent compensation based on the texture and color information around the shadows;

[0030] Illumination imbalance correction algorithm: Utilizes the characteristics of textiles, including fiber gloss and color distribution, to adaptively optimize image brightness and color uniformity, reducing color distortion caused by lighting changes;

[0031] Color and texture fidelity optimization: Combining color restoration and texture fidelity technology to ensure that the original color and texture characteristics of textiles are preserved while enhancing image contrast and brightness.

[0032] Furthermore, the adaptive local brightness adjustment is based on local contrast enhancement, using the following formula:

[0033] Among them, I enhanced(x, y) is the brightness of the enhanced image, I(x, y) is the brightness of the original image at position (x, y), α is the enhancement factor, C(x, y) is the local contrast of the original image at (x, y), It is the average local contrast of the image. The local contrast is enhanced by increasing the brightness of the area with a local contrast greater than the average and reducing the brightness of the area with a local contrast less than the average.

[0034] The intelligent shadow compensation mechanism is performed by the following steps:

[0035] Shadow detection, using thresholding method:

[0036] Where S(x, y) indicates whether the position (x, y) is in the shadow, L(x, y) is the brightness of the position (x, y), and T is the threshold for shadow detection;

[0037] The brightness compensation of the shadow area is expressed by the following formula: compensated (x, y) = I (x, y) + β × (TL (x, y)) × S (x, y), where, I compensated (x, y) is the brightness of the compensated image, and β is the compensation factor used to adjust the intensity of the compensation;

[0038] The illumination imbalance correction algorithm adopts a single-scale Retinex algorithm based on Retinex theory, and the calculation formula is as follows: R(x, y)=log(I(x, y))-log(F(x, y)*I(x, y)), where R(x, y) is the corrected image brightness, log is the natural logarithm, F(x, y) is the Gaussian filter kernel, and * represents the convolution operation.

[0039] Further, the introduction in S5 is based on the color difference evaluation analysis and specifically includes:

[0040] Color space conversion: convert multispectral images to CIELAB color space for color difference assessment;

[0041] Color difference calculation: Calculate the color difference of the image before and after processing in the CIELAB color space, using the formula: Among them, ΔL * , Δa * and Δb * Respectively represent the difference in brightness, green-red, and blue-yellow between the two images, color difference The larger the value, the lower the color fastness;

[0042] Consideration of fiber type and dye properties: Adjust the color difference results according to the fiber type of the textile (cotton, wool, silk, synthetic fiber) and the chemical properties of the dye used, taking into account the differences in the reaction of fibers and dyes to light, heat, and chemicals, and adjust the color difference value to reflect the influence of physical and chemical properties;

[0043] Analysis of the impact of test conditions: According to different test conditions (friction intensity, number of water washings, exposure time), the color difference results are weighted or corrected, and the correction coefficients or weights under different conditions are set to reflect the actual impact on color fastness;

[0044] Quantification and evaluation of color fastness: The color difference values ​​of all pixels are statistically analyzed, and the overall average color difference value is calculated as a quantitative indicator of color fastness. According to the correction results of fiber type, dye properties and test conditions, the overall color fastness score of the textile sample is given.

[0045] Furthermore, the color fastness quantification and evaluation specifically include:

[0046] Color difference adjustment: for each pixel Adjust according to fiber type and dye properties: Among them, F t is the fiber type adjustment factor, D p is the dye property adjustment factor;

[0047] Color fastness rating S overall : Adjust the color difference of all pixels Find the average and get the average color difference Calculate the comprehensive color fastness score based on the test condition weights: Among them, W c is the test condition weight, ∑W c is the sum of all test condition weights, is the weighted average of the adjusted color differences taking into account the weights of the test conditions.

[0048] Furthermore, the deep learning technology in S6 is based on a generative adversarial network, specifically including:

[0049] The conditional generative adversarial network cGAN is selected as the deep learning framework. The corresponding images are generated according to the given conditions (test conditions: washing, friction, exposure), and the cGAN model is trained to learn the potential changes of the multispectral images of textiles under specific test conditions. The input is the multispectral image of the textile, the condition is the test condition, and the output is the multispectral image after simulation processing. The objective function of CGAN is expressed as:

[0050] Among them, x is the real image, y is the conditional variable (test condition), and z is the value from the prior distribution p. z (z) is the noise vector, G is the generator, and D is the discriminator;

[0051] GANs training process: cGAN is trained by minimizing the loss function of the generator G and maximizing the loss function of the discriminator D. The loss function of the generator is expressed as:

[0052] The loss function of the discriminator is expressed as:

[0053]

[0054] Color attenuation feature extraction: The generator part of cGAN is used to extract color attenuation features, including local changes in color, texture changes, and pattern changes. The extracted high-level features are vectorized to obtain a high-dimensional feature vector, which represents the overall changes before and after image processing, and finally a comprehensive color fastness score is performed.

[0055] Furthermore, the color attenuation feature extraction specifically includes:

[0056] Feature extraction layer selection: In the generator network of CGAN, an intermediate layer is selected as the feature extraction layer, which can capture the high-level features of textile images, including texture, edges, and local color patterns, reflecting subtle changes in color attenuation;

[0057] Feature vectorization: Extract feature maps from the selected feature extraction layer and convert them into one-dimensional feature vectors. Let the feature map output by the feature extraction layer be F map , whose dimensions are H×W×D, where H, W, and D are height, width, and depth, respectively. Feature vectorization is achieved through average pooling or maximum pooling operations. The converted one-dimensional feature vector is expressed as : Among them, Pool(·) represents the pooling operation.

[0058] The comprehensive color fastness rating includes:

[0059] The high-level color attenuation feature vector extracted from cGAN Quantitative color fastness data S obtained from color difference evaluation 量化 The fusion is performed through feature splicing;

[0060] Comprehensive scoring model: Linear regression is used to learn the relationship between the fused features and the actual color fastness. The model input is the fused features, and the output is the predicted color fastness score S 预测 : in, represents the feature fusion operation, and MLModel(·) represents the linear regression model. .

[0061] A color fastness evaluation system, used to implement the above-mentioned color fastness evaluation method, includes the following modules:

[0062] Multispectral imaging module: uses a multispectral camera to capture images of textiles at different wavelengths to provide raw data for color fastness assessment;

[0063] Image preprocessing and enhancement module: Adopt adaptive image enhancement algorithm to preprocess multispectral images, including non-uniform illumination correction and shadow compensation;

[0064] Deep learning feature extraction and analysis module: Combined with conditional generative adversarial networks, it automatically extracts high-level features of textile images, including color attenuation and texture changes, to reflect the color fastness performance under different test conditions;

[0065] Color difference evaluation and quantification module: Compare the textile images before and after processing, calculate the color difference in CIELAB color space, quantify the degree of color change, and provide direct quantitative evaluation for color fastness;

[0066] Comprehensive evaluation and report generation module: Combines the advanced features extracted by the deep learning model and the quantitative data of color difference evaluation to comprehensively evaluate the color fastness of textiles.

[0067] Beneficial effects of the present invention:

[0068] The present invention, by combining the advanced color decay features extracted by generative adversarial networks (cGAN) and the quantitative data of traditional color difference evaluation, can more comprehensively capture the color changes of textiles under different test conditions. This comprehensive analysis method not only takes into account the basic color difference information, but also includes advanced features such as the texture, edge and local color pattern of the textile, thereby significantly improving the accuracy and reliability of color fastness evaluation. The advanced feature extraction using cGAN demonstrates a new way to analyze the color decay of textiles, which can automatically identify and learn the complex patterns and features that affect color fastness. It is not only suitable for color fastness evaluation under standardized test conditions, but also can adapt to more varied practical application scenarios, providing a more powerful analysis tool for textile design and quality control.

[0069] The present invention optimizes the color fastness evaluation process through automated deep learning technology, reducing the need for manual intervention and subjective judgment. This automated analysis method not only improves the evaluation efficiency, but also enables large-scale data analysis, providing efficient technical support for quality monitoring and product development in the textile industry, thereby optimizing resource utilization and reducing costs.

[0070] The present invention, combined with multispectral imaging technology, adaptive image enhancement algorithm and application based on color difference evaluation analysis, can achieve high-precision and comprehensive evaluation of the color fastness of textiles. Through multispectral imaging technology, the present invention can capture the subtle color changes of textiles in a wide spectrum range, providing a rich data basis for color fastness analysis. The adaptive image enhancement algorithm further improves the image quality, effectively solves problems such as non-uniform illumination and shadows, and ensures the accuracy of the analysis. The analysis method based on color difference evaluation uses these high-quality image data to directly evaluate the color fastness of textiles in a quantitative manner, so that the evaluation results are both accurate and easy to understand. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0072] Figure 1 A schematic diagram of the evaluation method flow in an embodiment of the present invention;

[0073] Figure 2 Schematic diagram of functional modules of an evaluation system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0074] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0075] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0076] like Figure 1 As shown, a color fastness evaluation method comprises the following steps:

[0077] S1: Set test conditions to simulate external factors encountered by textiles during actual use or processing, including setting corresponding conditions of extrusion, friction, washing, exposure and sweat stains;

[0078] S2: The textile samples are treated for a predetermined time under simulated test conditions to simulate actual wear conditions;

[0079] S3: Use multispectral imaging technology to obtain multispectral images of textiles and capture the color changes of textiles;

[0080] S4: Develop an adaptive image enhancement algorithm to preprocess the multispectral images from S3 to cope with the non-uniform lighting and shadow issues of textile images;

[0081] S5: Based on color difference evaluation analysis, the spectral information of multispectral images of textiles before and after treatment is compared to quantify color fastness, considering the effects of fiber type, dye properties and test conditions on color fastness;

[0082] S6: Apply deep learning technology to analyze the multispectral image after image enhancement, extract color attenuation features, and combine with the quantitative color fastness data in S5 to comprehensively evaluate the color fastness score;

[0083] S7: Develop visualization tools to present color fastness evaluation results in a graphical manner.

[0084] S1 specifically includes:

[0085] S11, setting of compression conditions: simulating the pressure on the textile during use, using standardized weights to compress the textile sample for a predetermined time, and the mass of the weights and the compression time are selected according to the actual use environment of the textile;

[0086] S12, setting of friction conditions: simulating the wear of textiles in daily use, repeatedly rubbing the surface of textile samples through a dedicated wear test instrument. The number and intensity of friction are set according to the standard test method to simulate the degree of wear in actual wearing or use;

[0087] S13, setting of washing conditions: simulating the washing fastness of textiles during the washing process by washing the textile samples multiple times in a standardized washing liquid. The number of washes, water temperature and type of detergent used are determined according to the material of the textiles and the expected use environment;

[0088] S14, setting of exposure conditions: simulating the light fastness of textiles under sunlight, by using ultraviolet lamps or exposing textile samples to natural light for a predetermined period to evaluate the effect of long-term or short-term exposure on the color of textiles. The exposure time and light intensity are adjusted according to the use scenario and light fastness requirements of the textiles;

[0089] S15, setting of sweat stain conditions: simulating the contact between textiles and human sweat. The textile samples are treated by using simulated sweat samples. The composition of the simulated sweat samples is prepared according to the standardized method to reflect the effect of sweat of different pH values ​​on the color of textiles. The treatment time and frequency simulate the conditions that may be encountered in actual wearing.

[0090] S3 specifically includes:

[0091] S31: Select a multispectral imaging system equipped with multiple filters of different wavelengths to ensure coverage of a wide band range from ultraviolet to near infrared, and capture textile images at multiple wavelengths to obtain the response of textiles under different spectra;

[0092] S32: Before imaging, ensure that the textile sample is flat and under standardized lighting conditions, and calibrate the imaging system using a standard color card to ensure the color accuracy of the acquired image;

[0093] S33: Imaging the textile sample at multiple predetermined wavelengths, collecting one image at each wavelength, and sequentially capturing images at each wavelength by changing the filters in the imaging system to construct a complete multispectral data set;

[0094] S34: De-noising, contrast adjustment and color correction are performed on the collected multi-wavelength images, and the processed images are synthesized into a complete multi-spectral image, which includes the response information of the textile in the entire spectral range.

[0095] S35: Identify color changes in textiles at different wavelengths by comparing multispectral images before and after treatment, with a particular focus on wavelengths that show significant changes that may indicate changes in color fastness.

[0096] The adaptive image enhancement algorithms in S4 specifically include:

[0097] Adaptive local brightness adjustment: An improved local brightness adjustment algorithm is used to automatically adjust the brightness and contrast according to the specific area of ​​the textile image, taking into account the texture and fiber structure of the textile, ensuring that the brightness adjustment process does not destroy the details of the textile or introduce unnatural visual effects;

[0098] Intelligent shadow compensation mechanism: Identify shadow areas in multispectral images and perform intelligent compensation based on the texture and color information around the shadows;

[0099] Illumination imbalance correction algorithm: Utilizes the characteristics of textiles, including fiber gloss and color distribution, to adaptively optimize image brightness and color uniformity, reducing color distortion caused by lighting changes;

[0100] Color and texture fidelity optimization: Combining color restoration and texture fidelity technology to ensure that the original color and texture characteristics of textiles are preserved while enhancing image contrast and brightness.

[0101] Adaptive local brightness adjustment is based on local contrast enhancement using the following formula:

[0102] Among them, I enhanced (x, y) is the brightness of the enhanced image, I(x, y) is the brightness of the original image at position (x, y), α is the enhancement factor, C(x, y) is the local contrast of the original image at (x, y), It is the average local contrast of the image. The local contrast is enhanced by increasing the brightness of the area with a local contrast greater than the average and reducing the brightness of the area with a local contrast less than the average.

[0103] The intelligent shadow compensation mechanism works through the following steps:

[0104] Shadow detection, using thresholding method:

[0105] Where S(x, y) indicates whether the position (x, y) is in the shadow, L(x, y) is the brightness of the position (x, y), and T is the threshold for shadow detection;

[0106] The brightness compensation of the shadow area is expressed by the following formula: compensated (x, y) = I (x, y) + β × (TL (x, y)) × S (x, y), where, I compensated (x, y) is the brightness of the compensated image, and β is the compensation factor used to adjust the intensity of the compensation;

[0107] The illumination imbalance correction algorithm adopts a single-scale Retinex algorithm based on Retinex theory. The calculation formula is as follows: R(x, y) = log(I(x, y)) - log(F(x, y)*I(x, y)), where R(x, y) is the corrected image brightness, log is the natural logarithm, F(x, y) is the Gaussian filter kernel, and * represents the convolution operation. Brightness correction is achieved by calculating the logarithmic difference between the image and its own convolution under Gaussian filtering.

[0108] The introduction of color difference evaluation analysis in S5 specifically includes:

[0109] Color space conversion: convert multispectral images to CIELAB color space for color difference assessment;

[0110] Color difference calculation: Calculate the color difference of the image before and after processing in the CIELAB color space, using the formula: Among them, ΔL *, Δa * and Δb * Respectively represent the difference in brightness, green-red, and blue-yellow between the two images, color difference The larger the value, the lower the color fastness;

[0111] Consideration of fiber type and dye properties: Adjust the color difference results according to the fiber type of the textile (such as cotton, wool, silk, synthetic fiber, etc.) and the chemical properties of the dyes used, consider the differences in the reactions of fibers and dyes to light, heat, and chemicals, and adjust the color difference values ​​to reflect the influence of physical and chemical properties;

[0112] Analysis of the impact of test conditions: According to different test conditions (such as friction intensity, number of water washings, exposure time, etc.), the color difference results are weighted or corrected, and the correction coefficients or weights under different conditions are set to reflect the actual impact on color fastness;

[0113] Quantification and evaluation of color fastness: The color difference values ​​of all pixels are statistically analyzed, and the overall average color difference value is calculated as a quantitative indicator of color fastness. According to the correction results of fiber type, dye properties and test conditions, the overall color fastness score of the textile sample is given.

[0114] The results of the color fastness assessment are organized into a report, including color difference statistics, correction factors, final color fastness score and other information, and charts or images are provided to show the color fastness changes so that the color changes can be intuitively understood.

[0115] Color fastness quantification and evaluation specifically include:

[0116] Color difference adjustment: for each pixel Adjust according to fiber type and dye properties: Among them, F t is the fiber type adjustment factor, D p is the dye property adjustment factor;

[0117] Color fastness rating S overall : Adjust the color difference of all pixels Find the average and get the average color difference Calculate the comprehensive color fastness score based on the test condition weights: Among them, W c is the test condition weight, ∑W c is the sum of all test condition weights, is the weighted average of the adjusted color differences taking into account the weights of the test conditions.

[0118] Adjustment factor F for fiber type t , synthetic fibers may have higher color fastness, so F tThe value is lower; for dye properties adjustment factor D p , natural dyes may have lower color fastness, so their D p The value is higher; for the test condition weight W c , different test conditions (such as washing, friction, exposure), assign a weight W c , to indicate the importance of the condition's effect on color fastness. For example, an exposure test may be given a higher weight than a washing test because light has a more significant effect on certain dyes.

[0119] The deep learning technology in S6 is based on generative adversarial networks, including:

[0120] The conditional generative adversarial network cGAN is selected as the deep learning framework. According to the given conditions (such as test conditions: washing, friction, exposure, etc.), the corresponding images are generated, and the cGAN model is trained to learn the potential changes of the multispectral images of textiles under specific test conditions. The input is the multispectral image of the textile, the condition is the test condition, and the output is the multispectral image after simulation processing. The objective function of CGAN is expressed as:

[0121] Among them, x is the real image, y is the conditional variable (test condition), and z is the value from the prior distribution p. z (z) is the noise vector, G is the generator, and D is the discriminator;

[0122] GANs training process: cGAN is trained by minimizing the loss function of the generator G and maximizing the loss function of the discriminator D. The loss function of the generator is expressed as:

[0123] The loss function of the discriminator is expressed as:

[0124]

[0125] Color attenuation feature extraction: The generator part of cGAN is used to extract color attenuation features, including local changes in color, texture changes, and pattern changes. The extracted high-level features are automatically learned through the internal layers of the generator network, and a high-dimensional feature vector is obtained. This high-dimensional feature vector represents the overall changes before and after the image is processed, and finally a comprehensive color fastness score is performed.

[0126] Color attenuation feature extraction specifically includes:

[0127] Feature extraction layer selection: In the generator network of CGAN, an intermediate layer is selected as the feature extraction layer, which can capture the high-level features of textile images, including texture, edges, and local color patterns, reflecting subtle changes in color attenuation;

[0128] Feature vectorization: Extract feature maps from the selected feature extraction layer and convert them into one-dimensional feature vectors. Let the feature map output by the feature extraction layer be F map , whose dimensions are H×W×D, where H, W, and D are height, width, and depth (number of channels), respectively. Feature vectorization is achieved through average pooling or maximum pooling operations. The converted one-dimensional feature vector is expressed as Among them, Poll(·) represents the pooling operation.

[0129] Comprehensive color fastness rating includes:

[0130] The high-level color attenuation feature vector extracted from cGAN Quantitative color fastness data S obtained from color difference evaluation 量化 The fusion is performed through feature splicing;

[0131] Comprehensive scoring model: Linear regression is used to learn the relationship between the fused features and the actual color fastness. The model input is the fused features, and the output is the predicted color fastness score S 预测 : in, represents the feature fusion operation, and MLModel(·) represents the linear regression model.

[0132] Through the above scheme, the generator part of cGAN can be more effectively utilized to extract high-level features about color attenuation, and the color fastness can be comprehensively evaluated through feature vectorization and fusion with traditional quantitative data, thereby providing a more comprehensive and in-depth color fastness analysis.

[0133] like Figure 2 As shown, a color fastness evaluation system is used to implement the above-mentioned color fastness evaluation method, including the following modules:

[0134] Multispectral imaging module: uses a multispectral camera to capture images of textiles at different wavelengths to provide raw data for color fastness assessment;

[0135] Image preprocessing and enhancement module: Adopt adaptive image enhancement algorithm to preprocess multispectral images, including non-uniform illumination correction and shadow compensation;

[0136] Deep learning feature extraction and analysis module: Combined with conditional generative adversarial networks, it automatically extracts high-level features of textile images, including color attenuation and texture changes, to reflect the color fastness performance under different test conditions;

[0137] Color difference evaluation and quantification module: Compare the textile images before and after processing, calculate the color difference in CIELAB color space, quantify the degree of color change, and provide direct quantitative evaluation for color fastness;

[0138] Comprehensive evaluation and report generation module: Combines the advanced features extracted by the deep learning model and the quantitative data of color difference evaluation to comprehensively evaluate the color fastness of textiles.

[0139] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present invention is limited to these examples. Under the concept of the present invention, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the present invention as described above, which are not provided in detail for the sake of simplicity.

[0140] The present invention is intended to cover all such substitutions, modifications and variations that fall within the broad scope of the claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for evaluating color fastness, characterized in that: The following steps are involved: S1: Set test conditions to simulate external factors encountered by textiles during actual use or processing, including setting corresponding conditions of extrusion, friction, washing, exposure and sweat stains; S2: The textile samples are treated for a predetermined time under simulated test conditions to simulate actual wear conditions; S3: Use multispectral imaging technology to obtain multispectral images of textiles and capture the color changes of textiles; S4: Develop an adaptive image enhancement algorithm to preprocess the multispectral images from S3 to cope with the non-uniform lighting and shadow issues of textile images; S5: Based on the color difference evaluation analysis, the spectral information of the multispectral images of the textiles before and after the treatment is compared to quantify the color fastness, taking into account the influence of fiber type, dye properties and test conditions on the color fastness, and the color difference evaluation analysis specifically includes: Color space conversion: convert multispectral images to CIELAB color space for color difference assessment; Color difference calculation: Calculate the color difference of the image before and after processing in the CIELAB color space, using the formula: Among them, ΔL * , Δa * and Δb * Respectively represent the difference in brightness, green-red and blue-yellow between the two images, color difference The larger the value, the lower the color fastness; Consideration of fiber type and dye properties: The color difference results are adjusted according to the fiber type of the textile and the chemical properties of the dye used. The color difference values ​​are adjusted to reflect the influence of physical and chemical properties, taking into account the differences in the reactions of fibers and dyes to light, heat, and chemicals; Analysis of the impact of test conditions: according to different test conditions, the color difference results are weighted or corrected, and the correction coefficient or weight under different conditions is set to reflect the actual impact on color fastness; Quantification and evaluation of color fastness: Statistical analysis is performed on the color difference values ​​of all pixels, and the overall average color difference value is calculated as a quantitative indicator of color fastness. Based on the correction results of fiber type, dye properties and test conditions, the overall color fastness score of the textile sample is given; The color fastness quantification and evaluation specifically include: Chromatic aberration adjustment: for each pixel Adjust according to fiber type and dye properties: Among them, F t is the fiber type adjustment factor, D p is the dye property adjustment factor; Color fastness rating S overall : Adjust the color difference of all pixels Find the average and get the average color difference Calculate the comprehensive color fastness score based on the test condition weights: Among them, W c is the test condition weight, ∑W c is the sum of all test condition weights, is the weighted average of the color differences adjusted to take into account the weights of the test conditions; S6: Apply deep learning technology to analyze the multispectral image after image enhancement, extract color attenuation features, and combine with the quantitative color fastness data in S5 to comprehensively evaluate the color fastness score. The deep learning technology is based on a generative adversarial network and specifically includes: The conditional generative adversarial network cGAN is selected as the deep learning framework. The corresponding images are generated according to the given conditions. The cGAN model is trained to learn the potential changes of the multispectral images of textiles under specific test conditions. The input is the multispectral image of the textile, the condition is the test condition, and the output is the multispectral image after simulation processing. The objective function of CGAN is expressed as: Among them, x is the real image, y is the conditional variable, and z is the variable from the prior distribution p. z (z) is the noise vector, G is the generator, and D is the discriminator; GANs training process: cGAN is trained by minimizing the loss function of the generator G and maximizing the loss function of the discriminator D. The loss function of the generator is expressed as: The loss function of the discriminator is expressed as: Color attenuation feature extraction: The generator part of cGAN is used to extract color attenuation features, including local changes in color, texture changes, and pattern changes. The extracted high-level features are vectorized to obtain a high-dimensional feature vector, which represents the overall changes before and after image processing, and finally a comprehensive color fastness score is performed; The color attenuation feature extraction specifically includes: Feature extraction layer selection: In the generator network of CGAN, an intermediate layer is selected as the feature extraction layer, which can capture the high-level features of textile images, including texture, edges, and local color patterns, reflecting subtle changes in color attenuation; Feature vectorization: Extract feature maps from the selected feature extraction layer and convert them into one-dimensional feature vectors. Let the feature map output by the feature extraction layer be F map , whose dimensions are H×W×D, where H, W, and D are height, width, and depth, respectively. Feature vectorization is achieved through average pooling or maximum pooling operations. The converted one-dimensional feature vector is expressed as : Among them, Pool(·) represents the pooling operation; The comprehensive color fastness rating includes: The high-level color attenuation feature vector extracted from cGAN Quantitative color fastness data S obtained from color difference evaluation 量化 The fusion is performed through feature splicing; Comprehensive scoring model: Linear regression is used to learn the relationship between the fused features and the actual color fastness. The model input is the fused features, and the output is the predicted color fastness score S 预测 : in, represents the feature fusion operation, and MLModel(·) represents the linear regression model; S7: Develop visualization tools to present color fastness evaluation results in a graphical manner.

2. A color fastness evaluation method according to claim 1, characterized in that: The S1 specifically includes: S11, setting of compression conditions: simulating the pressure on the textile during use, using standardized weights to compress the textile sample for a predetermined time; S12, setting of friction conditions: simulating the wear of textiles in daily use, and repeatedly rubbing the surface of textile samples through the wear test instrument; S13, setting of washing conditions: simulating the washing fastness of textiles during the cleaning process by washing the textile samples multiple times in a standardized washing liquid; S14, Setting of exposure conditions: Simulating the light fastness of textiles under sunlight, by using ultraviolet lamps or exposing textile samples to natural light for a predetermined period to evaluate the effect of long-term or short-term exposure on the color of textiles; S15, setting of sweat conditions: simulating the contact between textiles and human sweat, by treating the textile samples with simulated sweat samples.

3. A color fastness evaluation method according to claim 2, characterized in that: The S3 specifically includes: S31: Select a multispectral imaging system equipped with multiple filters of different wavelengths to ensure coverage of a wide band range from ultraviolet to near infrared, and capture textile images at multiple wavelengths to obtain the response of textiles under different spectra; S32: Before imaging, ensure that the textile sample is flat and under standardized lighting conditions, and calibrate the imaging system using a standard color card to ensure the color accuracy of the acquired image; S33: Imaging the textile sample at multiple predetermined wavelengths, collecting one image at each wavelength, and sequentially capturing images at each wavelength by changing the filters in the imaging system to construct a complete multispectral data set; S34: De-noising, contrast adjustment and color correction are performed on the collected multi-wavelength images, and the processed images are synthesized into a complete multi-spectral image, where the complete multi-spectral image includes the reaction information of the textile in the entire spectral range; S35: Identify color changes of textiles at different wavelengths by comparing multispectral images before and after processing.

4. A color fastness evaluation method according to claim 1, characterized in that: The adaptive image enhancement algorithm in S4 specifically includes: Adaptive local brightness adjustment: An improved local brightness adjustment algorithm is used to automatically adjust the brightness and contrast according to the specific area of ​​the textile image, taking into account the texture and fiber structure of the textile, ensuring that the brightness adjustment process does not destroy the details of the textile or introduce unnatural visual effects; Intelligent shadow compensation mechanism: Identify shadow areas in multispectral images and perform intelligent compensation based on the texture and color information around the shadows; Illumination imbalance correction algorithm: Utilizes the characteristics of textiles, including fiber gloss and color distribution, to adaptively optimize image brightness and color uniformity, reducing color distortion caused by lighting changes; Color and texture fidelity optimization: Combining color restoration and texture fidelity technology to ensure that the original color and texture characteristics of textiles are preserved while enhancing image contrast and brightness.

5. A color fastness evaluation method according to claim 4, characterized in that: The adaptive local brightness adjustment is based on local contrast enhancement using the following formula: Among them, I enhanced (x, y) is the brightness of the enhanced image, I(x, y) is the brightness of the original image at position (x, y), α is the enhancement factor, C(x, y) is the local contrast of the original image at (x, y), It is the average local contrast of the image. The local contrast is enhanced by increasing the brightness of the area with a local contrast greater than the average and reducing the brightness of the area with a local contrast less than the average. The intelligent shadow compensation mechanism is performed by the following steps: Shadow detection, using thresholding method: Where S(x, y) indicates whether the position (x, y) is in the shadow, L(x, y) is the brightness of the position (x, y), and T is the threshold for shadow detection; The brightness compensation of the shadow area is expressed by the following formula: compensated (x, y) = I (x, y) + β × (TL (x, y)) × S (x, y), where, I compensated (x, y) is the brightness of the compensated image, and β is the compensation factor used to adjust the intensity of the compensation; The illumination imbalance correction algorithm adopts a single-scale Retinex algorithm based on Retinex theory, and the calculation formula is as follows: R(x, y)=log(I(x, y))-log(F(x, y)*I(x, y)), where R(x, y) is the corrected image brightness, log is the natural logarithm, F(x, y) is the Gaussian filter kernel, and * represents the convolution operation.

6. A color fastness evaluation system, used to implement a color fastness evaluation method as claimed in any one of claims 1 to 5, characterized in that: Includes the following modules: Multispectral imaging module: uses a multispectral camera to capture images of textiles at different wavelengths to provide raw data for color fastness assessment; Image preprocessing and enhancement module: Adopt adaptive image enhancement algorithm to preprocess multispectral images, including non-uniform illumination correction and shadow compensation; Deep learning feature extraction and analysis module: Combined with conditional generative adversarial networks, it automatically extracts high-level features of textile images, including color attenuation and texture changes, to reflect the color fastness performance under different test conditions; Color difference evaluation and quantification module: Compare the textile images before and after processing, calculate the color difference in CIELAB color space, quantify the degree of color change, and provide direct quantitative evaluation for color fastness; The color fastness quantification and evaluation specifically include: Chromatic aberration adjustment: for each pixel Adjust according to fiber type and dye properties: Among them, F t is the fiber type adjustment factor, D p is the dye property adjustment factor; Color fastness rating S overall : Adjust the color difference of all pixels Find the average and get the average color difference Calculate the comprehensive color fastness score based on the test condition weights: Among them, W c is the test condition weight, ∑W c is the sum of all test condition weights, is the weighted average of the color differences adjusted to take into account the weights of the test conditions; Comprehensive evaluation and report generation module: Combines the advanced features extracted by the deep learning model and the quantitative data of color difference evaluation to comprehensively evaluate the color fastness of textiles.

Images