Textile state detection and analysis system

Abstract

The invention provides a textile state detection and analysis system, and relates to the field of state detection.The textile state detection and analysis system comprises the steps that firstly, a to-be-detected textile surface image is obtained and preprocessed, then a double-branch feature extraction network is constructed synchronously, and local edge features of microscopic flaws and a global distribution rule of a macrostructure are captured respectively; local details and global semantics are further associated through cross-scale feature fusion, multi-scale image coding features of the surface of the to-be-detected textile are generated, finally, defect positioning, classification and confidence evaluation are performed on the fused features in combination with a pre-trained deep learning model, and whether an alarm signal is sent out or not is automatically judged according to a detection result. Therefore, defect detection can still keep high sensitivity in a complex light environment, false detection and missing detection of a shadow or overexposure area are effectively inhibited, and the accuracy and reliability of defect detection are remarkably improved.

Description

Technical Field

[0001] The present application relates to the field of state detection, and more particularly, in an embodiment of the present application, to a textile state detection and analysis system. Background Art

[0002] In the field of textile manufacturing and quality control, textile condition detection is a key link to ensure product quality. During the production process, textiles may be affected by a variety of factors, such as mechanical damage, chemical contamination or human error, which may cause defects on the textile surface. Therefore, timely and accurate detection and identification of these defects is crucial to maintaining brand reputation and meeting market demand.

[0003] However, the dynamically changing lighting conditions in actual production environments (including fluctuations in natural light, uneven lighting from equipment, local shadows caused by fabric wrinkles, etc.) significantly interfere with the quality of captured images, making it difficult for traditional image processing algorithms to stably extract effective features. This light-sensitive defect is specifically manifested in the following ways: overexposure artifacts are easily generated in areas with strong reflections, details are lost in areas with low light levels, and features in shadow areas are confused with those of real defects. Existing technologies usually use fixed-threshold lighting compensation or single-scale feature extraction strategies, which can neither effectively eliminate the gradient distortion caused by non-uniform lighting nor coordinately analyze micro-texture and macro-structural features, resulting in low accuracy in defect detection in complex lighting environments.

[0004] Therefore, an optimized textile status detection and analysis solution is desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. An embodiment of the present application provides a textile state detection and analysis system, which first obtains the surface image of the textile to be inspected and performs image preprocessing, and then simultaneously constructs a dual-branch feature extraction network to capture the local edge features of microscopic defects and the global distribution law of the macroscopic structure respectively; further, cross-scale feature fusion is used to associate local details with global semantics, and multi-scale image coding features of the surface of the textile to be inspected are generated, and finally, the fused features are combined with a pre-trained deep learning model to realize defect location, classification and confidence assessment, and automatically determine whether to issue an alarm signal based on the detection results. In this way, defect detection can still maintain high sensitivity in complex light environments, effectively suppress false detection and missed detection in shadow or overexposed areas, thereby significantly improving the accuracy and reliability of defect detection.

[0006] According to one aspect of the present application, a textile state detection and analysis system is provided, comprising: A textile surface image acquisition module, used to acquire the surface image of the textile to be inspected acquired by a camera; A textile surface image enhancement module, used for performing image preprocessing on the textile surface image to be inspected to obtain an enhanced textile surface image to be inspected; A multi-scale feature encoding module, used for performing multi-scale image feature encoding on the enhanced surface image of the textile to be inspected to obtain multi-scale image encoding features of the surface of the textile to be inspected, wherein the multi-scale feature encoding module includes: a multi-scale feature extraction unit, used for extracting local detail information and global abstract information from the enhanced surface image of the textile to be inspected to obtain local detail encoding features of the surface of the textile to be inspected and global structure encoding features of the surface of the textile to be inspected; a multi-scale feature fusion unit, used for performing dynamic feature structure multi-scale fusion on the local detail encoding features of the surface of the textile to be inspected and the global structure encoding features of the surface of the textile to be inspected to obtain multi-scale image encoding features of the surface of the textile to be inspected; The alarm module is used to obtain the detection result and determine whether to issue an alarm signal based on the multi-scale image coding features of the surface of the textile to be detected.

[0007] Compared with the prior art, the textile state detection and analysis system provided by the present application first obtains the surface image of the textile to be inspected and performs image preprocessing, then simultaneously constructs a dual-branch feature extraction network to capture the local edge features of microscopic defects and the global distribution law of the macroscopic structure respectively; further, the local details and global semantics are associated through cross-scale feature fusion to generate multi-scale image coding features of the surface of the textile to be inspected, and finally, the fusion features are combined with the pre-trained deep learning model to realize defect location, classification and confidence evaluation, and automatically determine whether to issue an alarm signal based on the detection results. In this way, defect detection can still maintain high sensitivity in complex light environments, effectively suppress false detection and missed detection in shadow or overexposed areas, thereby significantly improving the accuracy and reliability of defect detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0009] Figure 1 4 is a system block diagram of a textile state detection and analysis system according to an embodiment of the present application.

[0010] Figure 2 Schematic diagram of data flow of a textile state detection and analysis system according to an embodiment of the present application.

[0011] Figure 34 is a block diagram of a multi-scale feature encoding module in a textile state detection and analysis system according to an embodiment of the present application.

[0012] Figure 4 It is a block diagram of a multi-scale feature fusion unit in a textile state detection and analysis system according to an embodiment of the present application.

[0013] Figure 5 It is a block diagram of a multi-scale image feature alignment and fusion subunit in a textile state detection and analysis system according to an embodiment of the present application.

[0014] Figure 6 A block diagram of a secondary subunit for attention weight calculation in a textile state detection and analysis system according to an embodiment of the present application. DETAILED DESCRIPTION

[0015] Various exemplary embodiments, features and aspects of the present application will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0016] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0017] In addition, in order to better illustrate the present application, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present application can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present application.

[0018] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0019] In the textile manufacturing process, the key to ensuring product quality lies in timely and accurate detection and identification of surface defects caused by mechanical damage, chemical contamination or human error. However, dynamically changing lighting conditions (such as natural light fluctuations, uneven equipment lighting and local shadows caused by fabric wrinkles) seriously affect the quality of image acquisition, making it difficult for traditional algorithms to stably extract effective features, thereby reducing the accuracy of defect detection. Existing technologies mostly rely on fixed-threshold lighting compensation and single-scale feature extraction strategies, which cannot effectively deal with gradient distortion caused by non-uniform lighting, and lack the ability to coordinate the analysis of micro-texture and macro-structure features, resulting in poor defect detection results in complex lighting environments.

[0020] Accordingly, the technical concept of the present application is to use image analysis and coding technology based on deep learning, first obtain the surface image of the textile to be inspected collected by the camera, then use the dynamic illumination correction algorithm to eliminate the interference of ambient light, and enhance the texture details of the surface image of the textile to be inspected based on adaptive noise suppression, and then simultaneously build a dual-branch feature extraction network to capture the local edge features of microscopic defects and the global distribution law of macroscopic structures respectively; further associate local details with global semantics through cross-scale feature fusion, generate multi-scale image coding features of the surface of the textile to be inspected, and finally combine the pre-trained deep learning model to realize defect location, classification and confidence assessment of the fused features, and automatically determine whether to issue an alarm signal based on the detection results. This solution can maintain high sensitivity of defect detection in complex light environments through dynamic illumination compensation and multi-scale feature collaborative analysis, effectively suppress false detection and missed detection in shadow or overexposed areas, thereby significantly improving the accuracy and reliability of defect detection.

[0021] In response to the above technical problems, the present application proposes a textile status detection and analysis system. Figure 1 4 is a system block diagram of a textile state detection and analysis system according to an embodiment of the present application. Figure 2 FIG. 1 is a data flow diagram of a textile state detection and analysis system according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to an embodiment of the present application, the textile state detection and analysis system 100 includes: a textile surface image acquisition module 110, which is used to obtain a surface image of the textile to be inspected acquired by a camera; a textile surface image enhancement module 120, which is used to perform image preprocessing on the surface image of the textile to be inspected to obtain an enhanced surface image of the textile to be inspected; a multi-scale feature encoding module 130, which is used to perform multi-scale image feature encoding on the enhanced surface image of the textile to be inspected to obtain multi-scale image encoding features of the surface of the textile to be inspected; and an alarm module 140, which is used to obtain a detection result and determine whether to issue an alarm signal based on the multi-scale image encoding features of the surface of the textile to be inspected.

[0022] In the above-mentioned textile state detection and analysis system 100, the textile surface image acquisition module 110 is used to obtain the surface image of the textile to be inspected collected by the camera. It should be understood that in the technical solution of the present application, the surface image of the textile to be inspected collected by the camera is obtained, and a high-resolution industrial camera is used for image acquisition. This type of camera has a high pixel density and color depth, and can carefully show the microscopic structural characteristics of the textile surface, such as fiber orientation, yarn gap and possible minor defects. At the same time, considering that the surface of the textile usually has a large area, the system is configured with an appropriate lens combination to meet the shooting requirements of textiles of different sizes, ensuring that the entire detection area can be clearly captured. It is worth mentioning that in order to further improve the accuracy and stability of image acquisition, the module integrates an automatic focusing function. By real-time monitoring of image clarity, the system can automatically adjust the camera lens position according to the specific morphology of the textile surface (such as flatness, thickness difference), ensuring that each shot can achieve the best focusing effect. At the same time, in response to the application requirements on high-speed production lines, the image acquisition module supports a fast continuous shooting mode, which can complete the image acquisition task of a large number of samples in a short time, meeting the rhythm requirements of efficient production.

[0023] In the above-mentioned textile state detection and analysis system 100, the textile surface image enhancement module 120 is used to perform image preprocessing on the surface image of the textile to be inspected to obtain an enhanced surface image of the textile to be inspected, including: performing noise suppression and uneven illumination correction on the surface image of the textile to be inspected to obtain the enhanced surface image of the textile to be inspected. It should be understood that, considering the dynamic illumination interference in the actual production environment (such as natural light fluctuations, uneven illumination of equipment, and shadow coverage caused by fabric wrinkles), the directly collected original image often contains a large amount of noise (such as sensor noise, motion blur) and non-uniform illumination distribution. For example, a strong reflective area may cover up the texture features of subtle defects due to overexposure, a weak light area may cause minor defects to be unrecognizable due to loss of details, and a shadow area is easily confused with the real defects. These problems seriously reduce the effectiveness of subsequent feature extraction, making it difficult for traditional detection algorithms to distinguish between real defects and artifacts caused by illumination or noise, ultimately leading to false detection or missed detection. If the illumination compensation method with a fixed threshold is difficult to adapt to the spatial heterogeneity of a complex light environment, a single-scale noise suppression strategy may also destroy the microtexture of the textile surface due to excessive smoothing, further weakening the feature resolution capability. Based on this, the present application performs noise suppression and illumination unevenness correction on the surface image of the textile to be inspected to obtain an enhanced surface image of the textile to be inspected. In particular, in a specific example of the present application, the illumination intensity distribution of each region in the image is analyzed in real time by a dynamic illumination correction algorithm, and an adaptive weight adjustment strategy is used to suppress the brightness of the overexposed area and to compensate the gradient of the shadow area, thereby eliminating the gradient distortion caused by non-uniform illumination; at the same time, combined with the adaptive noise suppression technology based on texture perception, the microstructural details of the textile surface (such as yarn gaps and fiber orientation) are retained while filtering out sensor noise and motion blur. This processing process not only overcomes the limitations of the fixed threshold method, but also ensures the consistency of the processing effect under different illumination conditions by dynamically adjusting parameters. That is, by eliminating illumination unevenness and noise interference, the enhanced image can clearly present the microtexture (such as broken yarn, hair balls) and macrostructure (such as wrinkles and color block distribution) of the textile surface. The image quality is significantly improved, laying a solid foundation for subsequent feature extraction and model reasoning.

[0024] In the above-mentioned textile state detection and analysis system 100, the multi-scale feature encoding module 130 is used to perform multi-scale image feature encoding on the enhanced surface image of the textile to be inspected to obtain multi-scale image encoding features of the surface of the textile to be inspected. It should be understood that, considering the wide variety of defects on the surface of textiles, there are both local defects such as broken yarn and hair balls at the micro level, and structural abnormalities such as wrinkles and uneven distribution of color blocks at the macro level. Single-scale feature extraction methods are often difficult to fully describe these complex defect types and their spatial distribution patterns. By performing multi-scale image feature encoding on the enhanced surface image of the textile to be inspected to obtain multi-scale image encoding features of the surface of the textile to be inspected, multi-level information in the image can be captured, and the accuracy and robustness of defect detection can be improved.

[0025] Figure 3 FIG. 1 is a block diagram of a multi-scale feature encoding module in a textile state detection and analysis system according to an embodiment of the present application. Figure 3 As shown, the multi-scale feature encoding module 130 includes: a multi-scale feature extraction unit 131, which is used to extract local detail information and global abstract information from the enhanced surface image of the textile to be inspected to obtain local detail coding features of the textile surface to be inspected and global structure coding features of the textile surface to be inspected; a multi-scale feature fusion unit 132, which is used to perform dynamic feature structure multi-scale fusion on the local detail coding features of the textile surface to be inspected and the global structure coding features of the textile surface to be inspected to obtain the multi-scale image coding features of the textile surface to be inspected.

[0026] In an embodiment of the present application, the multi-scale feature extraction unit 131 is used to: use a multi-scale feature extractor based on a hollow space pyramid pooling model to extract local detail information and global abstract information from the enhanced surface image of the textile to be inspected to obtain a local detail coding feature map of the surface of the textile to be inspected as the local detail coding feature of the surface of the textile to be inspected and a global structure coding feature map of the surface of the textile to be inspected as the global structure coding feature of the surface of the textile to be inspected. It should be understood that, considering that textile surface defects (such as broken yarn, hair balls, fine stains, etc.) are usually small in size and complex in texture, their edges and morphological features may only occupy a few pixels in width in the image, and are easy to form visual confusion with normal textile textures (such as yarn interlacing structure, fiber orientation). Especially in the enhanced image after dynamic illumination correction, although the problems of noise and uneven illumination are alleviated, the local features of tiny defects still need to be accurately characterized by highly sensitive feature extraction technology. At the same time, considering that the defects on the surface of textiles include not only microscopic local defects such as broken yarns and pilling, but also macroscopic structural abnormalities such as large-scale wrinkles, uneven distribution of color blocks or pattern dislocation, the identification of these problems depends on the understanding of the overall semantics of the image, such as large-area chromatic aberration defects need to be judged in combination with the color distribution law of the entire image. Therefore, the present application extracts local detail information and global abstract information from the enhanced surface image of the textile to be inspected to obtain local detail coding features of the surface of the textile to be inspected and global structural coding features of the surface of the textile to be inspected. In particular, in a specific example of the present application, a multi-scale feature extractor based on a hollow space pyramid pooling model is used to extract local detail information and global abstract information from the enhanced surface image of the textile to be inspected to obtain a local detail coding feature map of the surface of the textile to be inspected as the local detail coding features of the surface of the textile to be inspected and a global structural coding feature map of the surface of the textile to be inspected as the global structural coding features of the surface of the textile to be inspected. Specifically, the model deploys multiple dilated convolution layers with different dilation rates on a shared input feature map, and simultaneously captures the high-frequency edge features of tiny defects (such as focusing on pixel-level gradient changes with small dilation rate convolution) and the context-related features of large-scale defects (such as covering a wider area with large dilation rate convolution to identify continuous defects) without reducing the spatial resolution. For example, for the common tiny pilling defects on the surface of textiles, small dilation rate convolution can accurately capture its local clustered gradient distribution; and for long scratches spanning multiple yarn cycles, large dilation rate convolution can effectively integrate its linear extension mode, thereby improving the sensitivity and characterization capabilities of multi-scale defects.

[0027] Specifically, the multi-scale feature fusion unit 132 is used to perform dynamic feature structure multi-scale fusion on the local detail coding features of the surface of the textile to be inspected and the global structure coding features of the surface of the textile to be inspected to obtain the multi-scale image coding features of the surface of the textile to be inspected. It should be understood that, considering that in the textile state detection and analysis system, the defect detection of the textile surface needs to take into account the coordinated analysis of microscopic local anomalies (such as broken yarn, hair balls) and macroscopic structural distortions (such as wrinkles, color difference), however, the traditional method is prone to misjudgment due to the misalignment of the spatial geometric structure of local features and global features (such as fabric wrinkles causing local texture deformation) and the fragmentation of semantic information (such as the disconnection of local color difference and global color distribution law). For example, the local detail feature may accurately capture the sharp edge of a tiny broken yarn, but it is impossible to determine whether it is an artifact under the shadow of the wrinkle; although the global feature can identify the wavy distortion of a large range of wrinkles, it is difficult to locate the exact position of the tiny defect. Traditionally, simple feature splicing or weighted averaging strategies are used, which can neither solve the spatial alignment problem of cross-scale features nor model the complex interactions between features, resulting in redundant feature expressions after fusion and sensitivity to light interference. Based on this, the present application obtains multi-scale image coding features of the surface of the textile to be inspected by performing dynamic feature structure multi-scale fusion on the local detail coding features of the surface of the textile to be inspected and the global structure coding features of the surface of the textile to be inspected. In particular, this method decouples local and global feature maps into multiple groups of local coding matrices, uses feature phase similarity metrics to guide the dynamic search alignment process, and finally generates multi-scale image coding features through attention-driven saliency aggregation.

[0028] Figure 4 FIG. 1 is a block diagram of a multi-scale feature fusion unit in a textile state detection and analysis system according to an embodiment of the present application. Figure 4 As shown, in an embodiment of the present application, the multi-scale feature fusion unit 132 includes: a feature decoupling subunit 1321 for the textile to be inspected, which is used to perform feature decoupling on the local detail coding feature map of the textile surface to be inspected and the global structure coding feature map of the textile surface to be inspected to obtain a set of local detail feature coding matrices of the textile surface to be inspected and a set of global structure local feature coding matrices of the textile surface to be inspected; a multi-scale image feature alignment and fusion subunit 1322, which is used to perform phase alignment and attention saliency fusion on the set of local detail feature coding matrices of the textile surface to be inspected and the set of global structure local feature coding matrices of the textile surface to be inspected to obtain a multi-scale image coding feature map of the textile surface to be inspected as the multi-scale image coding feature of the textile surface to be inspected.

[0029] Specifically, the feature decoupling subunit 1321 of the textile to be inspected is used to perform feature decoupling on the local detail coding feature map of the textile to be inspected surface and the global structure coding feature map of the textile to be inspected surface to obtain a set of local detail feature coding matrices of the textile to be inspected surface and a set of local feature coding matrices of the global structure of the textile to be inspected surface, which is expressed as: in, is a local detail coding feature map of the surface of the textile to be inspected, is the global structural coding feature map of the surface of the textile to be inspected, To perform feature decoupling operations, , , and They are the first, second, and third in the set of local detail feature coding matrices of the textile surface to be inspected. and A local detail feature coding matrix of the textile surface to be inspected, , , and They are the first, second, and third local feature encoding matrices of the global structure of the textile surface to be inspected. and The global structure local feature coding matrix of the surface of the textile to be inspected. It should be understood that considering that the detection of the textile surface needs to analyze both microscopic local defects (such as broken yarn, hair balls) and macroscopic structural abnormalities (such as wrinkles, color difference), there are significant differences in the spatial geometric structure between the local detail features (focusing on pixel-level high-frequency information) and the global structure features (representing long-range semantic associations): local features may cause spatial deformation due to fabric wrinkles (such as yarn distortion), and global features may lose the precise position of tiny defects due to downsampling. If high-dimensional features are processed directly, there will be problems such as high computational complexity, semantic confusion caused by local-global feature space misalignment (such as misjudging local breaks in wrinkle shadows as defects), and difficulty in adapting to non-uniform deformation caused by dynamic lighting due to the strong integrity of the feature map. Based on this, the local detail coding feature map of the textile surface to be inspected and the global structure coding feature map of the textile surface to be inspected are feature decoupled. In particular, feature decoupling transforms global features into independently operable local semantic units through structured decomposition technology, which not only reduces the computational complexity of high-dimensional features, but also provides an operational basis for granular adaptation for subsequent cross-scale feature alignment by preserving the spatial correlation between local and global features. This decoupling is not a simple segmentation, but actually constructs an intermediate representation form of local and global features, so that the subsequent fusion stage can more accurately associate the local anomalies of microscopic defects with the overall semantics of the macroscopic structure, thereby improving the robustness of the detection system under complex working conditions.

[0030] Figure 5 FIG. 1 is a block diagram of a multi-scale image feature alignment and fusion subunit in a textile state detection and analysis system according to an embodiment of the present application. Figure 5As shown, in an embodiment of the present application, the multi-scale image feature alignment fusion subunit 1322 includes: a feature phase dynamic search alignment secondary subunit 1322-1, which is used to perform feature phase dynamic search alignment on the set of local detail feature coding matrices of the textile surface to be inspected and the set of global structure local feature coding matrices of the textile surface to be inspected based on the feature phase alignment between any two local detail feature coding matrices of the textile surface to be inspected and the local feature coding matrices of the textile surface to be inspected in the set of local detail feature coding matrices of the textile surface to be inspected and the set of global structure local feature coding matrices of the textile surface to be inspected to obtain a set of phase-aligned {local detail feature coding matrices of the textile surface to be inspected, local feature coding matrices of the textile surface to be inspected} feature pairs; an attention weight calculation secondary subunit 1322-1. 22-2, used to input each phase-aligned {local detail feature coding matrix of the textile surface to be tested, local feature coding matrix of the global structure of the textile surface to be tested} feature pair in the set of phase-aligned {local detail feature coding matrix of the textile surface to be tested, local feature coding matrix of the global structure of the textile surface to be tested} feature pairs into a feature joint perceptual attention network to obtain a set of joint perceptual attention weights of the textile surface features to be tested; the multi-scale salient aggregation secondary subunit 1322-3 of the textile surface to be tested, used to perform attention-driven salient aggregation on the set of phase-aligned {local detail feature coding matrix of the textile surface to be tested, local feature coding matrix of the global structure of the textile surface to be tested} feature pairs based on the set of joint perceptual attention weights of the textile surface features to be tested to obtain the multi-scale image coding feature map of the textile surface to be tested.

[0031] Specifically, the characteristic phase dynamic search and alignment secondary subunit 1322-1 is used to perform characteristic phase dynamic search and alignment on the set of local detail feature coding matrices of the textile surface to be inspected and the set of global structure local feature coding matrices of the textile surface to be inspected based on the characteristic phase alignment between any two local detail feature coding matrices of the textile surface to be inspected and the local feature coding matrices of the textile surface to be inspected in the set of local detail feature coding matrices of the textile surface to be inspected and the set of global structure local feature coding matrices of the textile surface to be inspected to obtain a set of phase-aligned {local detail feature coding matrices of the textile surface to be inspected, local feature coding matrices of the textile surface global structure to be inspected} feature pairs, which is expressed as a phase dynamic search and alignment formula: in, is the transpose operation, To calculate the Frobenius norm, yes and The characteristic phase alignment between To return the maximum value value, The method is to find the position of the maximum approximate matching value in the set of local feature encoding matrices of the global structure of the textile surface to be inspected. It should be understood that due to the problems of wrinkle deformation and dynamic uneven illumination during the processing of textiles, there are significant non-rigid deformation and phase shift between local detail features (such as the sharp edge of broken yarn) and global structure features (such as the wavy distribution of wrinkles) at the spatial geometric level. The traditional feature alignment method based on physical position matching is easily disturbed by such deformation, resulting in the failure of cross-scale feature association, which in turn leads to false detection or missed detection. For example, the local broken yarn feature may be offset in the global structure due to local stretching of the fabric. If it directly relies on coordinate alignment, its semantic association will be broken. To this end, by quantifying the structural similarity in the feature space (rather than the physical position correspondence), a sliding window is used to traverse the set of local detail feature encoding matrices of the textile surface to be inspected and the set of local feature encoding matrices of the global structure of the textile surface to be inspected, and the F-norm maximization matching mechanism is used to dynamically search for the best alignment relationship to establish a high-confidence cross-scale semantic association. This matrix norm-based similarity measure can capture the essential correlation between local and global features in texture gradient distribution, rather than simply relying on pixel-level position matching, thereby effectively overcoming the spatial misalignment caused by fabric deformation and illumination interference. In this way, the best feature pairs are adaptively matched through a dynamic search mechanism, and a high-confidence cross-scale semantic association is established. In addition, through the global sensitivity of the F-norm to the global feature distribution, the dynamic search mechanism driven by the maximum value of the F-norm enhances the robustness of the system to local deformation and illumination changes, overcomes the spatial misalignment caused by fabric deformation and illumination interference, and provides a high-quality multi-scale feature alignment foundation for the subsequent fusion stage.

[0032] Figure 6 FIG. 1 is a block diagram of a secondary subunit for calculating attention weight in a textile state detection and analysis system according to an embodiment of the present application. Figure 6As shown, in an embodiment of the present application, the attention weight calculation secondary subunit 1322-2 includes: a tertiary subunit 1322-21 for calculating scale association of the textile surface to be inspected, which is used to calculate the association matrix between each phase-aligned {local detail feature coding matrix of the textile surface to be inspected, local feature coding matrix of the textile surface global structure} feature pair in the set of the phase-aligned {local detail feature coding matrix of the textile surface to be inspected, local feature coding matrix of the textile surface global structure} feature pairs to obtain a set of scale association matrices of the textile surface to be inspected; a tertiary subunit 1322-22 for measuring the trace of the textile surface to be inspected, which is used to measure the scale association matrices of each textile surface to be inspected in the set of scale association matrices of the textile surface to be inspected to obtain a set of trace measurement values ​​of the textile surface to be inspected; a tertiary subunit 1322-23 for normalizing the trace measurement weights, which is used to obtain a set of joint perceptual attention weights of the surface features of the textile to be inspected based on the set of trace measurement values ​​of the textile surface to be inspected. It should be understood that although the semantic association between local details and global structural features is established through the phase dynamic search alignment mechanism, there are significant differences in the contribution of different feature pairs in actual scenes. For example, some feature pairs may contain complementary information of key defects (such as the association between the high-frequency gradient of broken yarn and the global yarn orientation deviation), while others may introduce misleading signals due to residual noise of illumination or wrinkle texture (such as local-global feature pairs in the wrinkle shadow area). Traditional methods use fixed weights or simple averaging strategies, which makes it difficult to distinguish effective features from noise, resulting in fusion results that are sensitive to environmental interference. In the technical solution of the present application, by introducing a feature joint perception attention network, dynamic weight allocation is performed on the feature pairs after phase alignment to solve the redundant interference problem that still exists after cross-scale feature alignment. That is, the present application uses a feature joint perception attention network to deeply explore the complex interaction patterns between phase-aligned feature pairs, learn to dynamically allocate weights, suppress artifact feature pairs caused by interference such as shadows and wrinkles, and highlight the cross-scale feature interaction patterns of real defects. Its essence is to learn the complex interaction patterns between feature pairs through deep learning models, and identify which cross-scale feature associations can effectively represent real defects (such as the strong correlation between local gradient aggregation in the ball area and global yarn density anomalies) and which associations only reflect environmental noise (such as artifacts in the shadow area of ​​wrinkles).

[0033] Specifically, the processing process of the textile surface pair scale association calculation tertiary subunit 1322-21 and the textile surface trace measurement tertiary subunit 1322-22 is expressed as follows: in, is the trace value of the matrix, is a phase-aligned {local detail feature encoding matrix of the textile surface to be inspected, local feature encoding matrix of the global structure of the textile surface to be inspected} feature pair, yes and The surface trace value of the textile to be tested is between .

[0034] Specifically, the trace measurement value weight normalization three-level sub-unit 1322-23 is used to: based on the set of trace measurement values ​​of the textile surface to be inspected, perform matrix manifold optimization based on the number of dual connections on each phase-aligned {local detail feature coding matrix of the textile surface to be inspected, local feature coding matrix of the textile surface global structure} feature pair in the set of phase-aligned {local detail feature coding matrix of the textile surface to be inspected, local feature coding matrix of the textile surface global structure} feature pairs to obtain a set of phase-optimized aligned {local detail feature coding matrix of the textile surface to be inspected, local feature coding matrix of the textile surface global structure} feature pairs; based on the set of phase-optimized aligned {local detail feature coding matrix of the textile surface to be inspected, local feature coding matrix of the textile surface global structure} feature pairs, obtain a set of joint perceptual attention weights of the surface features of the textile to be inspected.

[0035] More specifically, based on the set of surface trace values ​​of the textile to be tested, each phase-aligned {local detail feature coding matrix of the textile to be tested surface, local feature coding matrix of the global structure of the textile to be tested surface} feature pair in the set of phase-aligned {local detail feature coding matrix of the textile to be tested surface, local feature coding matrix of the global structure of the textile to be tested surface} feature pairs is subjected to matrix manifold optimization based on the number of dual connections to obtain a set of phase-optimized aligned {local detail feature coding matrix of the textile to be tested surface, local feature coding matrix of the global structure of the textile to be tested surface} feature pairs, and the specific processing process is as follows: In particular, here, for each matrix trace metric operation and ,make and ,and , , by calculating the corresponding and Between distance and distance The number of eigenvalues ​​of the matrix and Number of dual connections between and , that is, the distribution map stationarity between the phase-aligned local detail feature coding matrix of the textile surface to be inspected and the global structure local feature coding matrix of the textile surface to be inspected.

[0036] Then, in the discrete manifold representation of the local detail feature coding matrix of the textile surface to be inspected and the local feature coding matrix of the global structure of the textile surface to be inspected, the trace metric operation of the local detail feature coding matrix of the textile surface to be inspected and the local feature coding matrix of the global structure of the textile surface to be inspected is embedded and optimized by compactification of the embedding manifold based on the number of dual connections, so as to improve the calculation accuracy of the joint matrix trace metric on the basis of the robustness of long-range correlation connectivity: That is, based on the initial and Calculated , and In the case of , , ,initial and Optimize and then calculate Compute the optimized trace metric.

[0037] More specifically, based on the phase optimization alignment of the set of {local detail feature encoding matrix of the surface of the textile to be inspected, local feature encoding matrix of the surface of the textile to be inspected} feature pairs, a set of joint perceptual attention weights of the surface features of the textile to be inspected is obtained, which is expressed by the formula: in, Calculate the feature joint perception attention weight, is the phase-optimized alignment {local detail feature encoding matrix of the textile surface to be inspected, local feature encoding matrix of the global structure of the textile surface to be inspected} feature pair, yes The optimized phase optimization aligns the local detail feature coding matrix of the textile surface to be inspected. yes The optimized phase optimization aligns the local feature encoding matrix of the global structure of the textile surface to be inspected. yes and The optimized surface trace value of the textile to be tested is is the normalization function, for and The joint perceptual attention weights of the surface features of the textile to be inspected.

[0038] Specifically, the multi-scale significant aggregation secondary subunit 1322-3 of the textile surface to be inspected is used to perform attention-driven significant aggregation on the set of phase-aligned {local detail feature encoding matrix of the textile surface to be inspected, local feature encoding matrix of the global structure of the textile surface to be inspected} feature pairs based on the set of joint perceptual attention weights of the textile surface to be inspected to obtain the multi-scale image encoding feature map of the textile surface to be inspected, which is expressed as: in, and are the first joint perception weight matrix and the second joint perception weight matrix, It is subtracted by position point. It is added by location point. yes and The combined surface feature joint perception matrix of the textile to be inspected is the first one in the set of the surface feature joint perception matrix of the textile to be inspected. A joint perception matrix of the surface features of the textile to be inspected. , and They are the first, second, and third in the set of joint perception matrices of the surface features of the textile to be inspected. The joint perception matrix of the surface features of the textile to be inspected, It is the multi-scale image coding feature map of the surface of the textile to be inspected. It should be understood that in the technical solution of the present application, by introducing the attention-driven significant aggregation mechanism, the weight distribution of different feature pairs is dynamically adjusted. Its essence is to learn the complex interaction patterns between feature pairs through a deep learning model, identify which cross-scale feature associations can effectively characterize real defects (such as the strong correlation between the local gradient aggregation in the hair ball area and the global yarn density anomaly), and which associations only reflect environmental noise (such as artifacts in the wrinkle shadow area). Specifically, the system multiplies the phase-aligned {local detail feature coding matrix of the surface of the textile to be inspected, local feature coding matrix of the global structure of the surface of the textile to be inspected} feature pairs (such as the local gradient matrix representing the broken yarn defect and the global matrix reflecting the overall deviation of the yarn direction) with their corresponding attention weights, strengthen the expression of high-weight feature pairs (such as cross-scale synergistic signals of real defects), and suppress low-weight feature pairs (such as local-global artifact associations in the wrinkle shadow area). Among them, for high-weight feature pairs (such as local and global synergistic features of real defects), the synergistic effect is amplified by adding operations at position points to strengthen the cross-scale representation of defects; for low-weight feature pairs (such as interference signals in wrinkle areas), their negative effects are suppressed by subtracting operations at position points. This mechanism enables the system to adaptively filter interference signals caused by uneven lighting or fabric deformation while retaining key defect features, and ultimately generates a multi-scale image encoding feature map of the surface of the textile to be inspected with high discrimination. In this way, the system's adaptability to complex working conditions is significantly improved. For example, when detecting overexposed areas, the attention network assigns a lower weight to the features of this area to reduce the interference of distorted local features on the fusion results, while enhancing the weights of adjacent unaffected areas, using the integrity of global structural features to compensate for local information loss; for real defects, high weights are used to focus on the coordinated expression of local anomalies and global semantics (such as the strong correlation between local gradient aggregation in the ball area and global yarn density distribution anomalies), forming a significantly enhanced defect feature response, thereby adaptively highlighting the cross-scale representation of key defects, thereby maintaining high sensitivity and accuracy under complex lighting and deformation conditions.

[0039] In the above-mentioned textile state detection and analysis system 100, the alarm module 140 is used to obtain the detection result and determine whether to issue an alarm signal based on the multi-scale image coding features of the surface of the textile to be inspected, and is used to: input the multi-scale image coding feature map of the surface of the textile to be inspected into the trained defect detection model to obtain the detection result; based on the detection result, determine whether to issue an alarm signal.

[0040] Specifically, the multi-scale image coding feature map of the surface of the textile to be inspected is input into the trained defect detection model to obtain the detection result. In the embodiment of the present application, the detection result includes the position of the defect point, the type of defect point and the confidence. That is, the multi-scale image coding feature map of the surface of the textile to be inspected obtained by multi-scale feature fusion of the local detail coding feature map of the surface of the textile to be inspected and the global structure coding feature map of the surface of the textile to be inspected is used for detection to obtain the detection result. In this way, the position of the defect point can be accurate to the pixel level (such as the sub-pixel coordinates of the hole boundary), the defect type can be fine-grainedly classified based on the semantic association of cross-scale features (such as distinguishing between broken yarn caused by mechanical damage and color spots formed by chemical pollution), and the confidence reflects the credibility of the detection result through the probability distribution output by the model (such as high confidence marks clear defects, and low confidence prompts manual re-inspection). It is worth mentioning that in the technical solution of the present application, the defect detection model that has been trained is constructed based on deep learning technology, aiming to achieve efficient and accurate identification of textile surface defects, involving multiple key steps, including data input, model design and training, data enhancement, model optimization and final data output. Specifically, first, the data input stage integrates the feature map data stream of multi-scale features and high-quality defect annotation data sets. These data sets contain defect samples and non-defect samples of various types, forms and lighting conditions, ensuring that the model can fully cover various situations that may be encountered in actual production. In this way, the model can not only learn different types of defect features, but also adapt to different lighting and background conditions, thereby improving its robustness in complex environments. Then, in the model design and training stage, the target detection model selected is usually a target detection framework such as YOLOv5 or Faster R-CNN. These frameworks can directly locate and identify the location and category of defects in images, with high detection accuracy and speed. In addition, semantic segmentation models such as U-Net or DeepLabv3+ are also used to accurately segment the pixel-level area of ​​defects, further refine the detection results, and ensure that each defect can be accurately captured. In order to improve the generalization and robustness of the model, data enhancement technology is further used. Through operations such as image rotation, flipping, scaling, cropping, and illumination transformation, a large number of variant images can be generated to expand the training data set. This approach not only increases the training sample size of the model, but also enables the model to maintain stable performance in the face of unknown or changing environments. Next, use appropriate loss functions (such as Focal Loss, Dice Loss) and optimizers (such as AdamW), and continuously adjust the model parameters through model parameter adjustment and optimization to achieve the best performance. For example, Focal Loss can effectively solve the problem of imbalance between positive and negative samples, while Dice Loss helps to improve the model's ability to detect small targets.At the same time, the model inference speed is accelerated by hardware such as GPU to ensure efficient real-time detection. In this way, a high-performance defect detection model is finally obtained. The model can collect and pre-process image data streams in real time and input them into the trained model for real-time defect recognition and classification. The model inference speed is accelerated by hardware such as GPU to meet the real-time requirements of high-speed production lines. The results of the model output include the location coordinates, category labels, and confidence information of the defects, which will be used in subsequent links.

[0041] Specifically, based on the detection results, it is determined whether to issue an alarm signal. It should be understood that by real-time monitoring of the defect detection results, once the defect point location and defect point type are detected and a certain confidence level is reached, indicating that there is a problem with the textile, the system immediately issues an alarm signal. This can remind the on-site operator in the first time, so that they can quickly check and debug the production equipment, correct the abnormalities in the production process in time, and control the production of defective textiles to a minimum range. For example, when the system identifies a yarn break defect located in the central area of ​​the fabric, its specific coordinates are (X1, Y1) to (X2, Y2), and the type is identified as "yarn break caused by mechanical damage", with a confidence level of up to 95%. Based on the above detection results, the system will comprehensively consider the location, type and confidence level of the defect to make a final decision to determine whether to issue an alarm signal. Among them, since the defect is located in the core visual area of ​​the fabric and is a key type that affects product quality, the system decides to trigger the alarm mechanism. The alarm signal is first conveyed to the operator through the sound and light alarm installed on site. The alarm emits a harsh sound and a flashing red light, which quickly attracts the attention of the production line staff. This instant visual and auditory prompt can remind relevant personnel to take immediate action to prevent the problem from spreading further. At the same time, the system sends detailed alarm information to the mobile device of the production line supervisor through network communication, notifying him to go to the problem area for inspection immediately. The alarm information not only includes the specific location, type and confidence of the defect, but also comes with real-time image clips to quickly locate and analyze the root cause of the problem. This intelligent monitoring and early warning system greatly improves production stability and product quality, and provides a strong guarantee for the realization of automated and intelligent quality control.

[0042] In summary, the textile state detection and analysis system 100 based on the embodiment of the present application is explained, which first obtains the surface image of the textile to be inspected and performs image preprocessing, and then simultaneously constructs a dual-branch feature extraction network to capture the local edge features of microscopic defects and the global distribution law of macroscopic structures respectively; further, cross-scale feature fusion is used to associate local details with global semantics, and multi-scale image coding features of the surface of the textile to be inspected are generated, and finally, the fusion features are combined with the pre-trained deep learning model to realize defect location, classification and confidence evaluation, and automatically determine whether to issue an alarm signal based on the detection results. In this way, defect detection can still maintain high sensitivity in complex light environments, effectively suppress false detection and missed detection in shadow or overexposed areas, thereby significantly improving the accuracy and reliability of defect detection.

[0043] As described above, the textile state detection and analysis system 100 according to the embodiment of the present application can be implemented in various terminal devices. In one example, the textile state detection and analysis system 100 can be integrated into the terminal device as a software module and / or a hardware module. For example, the textile state detection and analysis system 100 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the textile state detection and analysis system 100 can also be one of the many hardware modules of the terminal device.

[0044] Alternatively, in another example, the textile state detection and analysis system 100 and the terminal device may also be separate devices, and the textile state detection and analysis system 100 may be connected to the terminal device via a wired and / or wireless network, and transmit interactive information in an agreed data format.

[0045] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0046] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0047] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0048] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0049] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

[0050] In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the system claim can also be implemented by one unit or device through software or hardware. The second and other words are used to indicate names, but not to indicate any particular order.

[0051] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit of the technical solution of the present invention.

Claims

1. A textile state detection and analysis system, characterized in that: include: A textile surface image acquisition module, used to acquire the surface image of the textile to be inspected acquired by a camera; A textile surface image enhancement module, used for performing image preprocessing on the textile surface image to be inspected to obtain an enhanced textile surface image to be inspected; A multi-scale feature encoding module, used for performing multi-scale image feature encoding on the enhanced surface image of the textile to be inspected to obtain multi-scale image encoding features of the surface of the textile to be inspected, wherein the multi-scale feature encoding module includes: a multi-scale feature extraction unit, used for extracting local detail information and global abstract information from the enhanced surface image of the textile to be inspected to obtain local detail encoding features of the surface of the textile to be inspected and global structure encoding features of the surface of the textile to be inspected; a multi-scale feature fusion unit, used for performing dynamic feature structure multi-scale fusion on the local detail encoding features of the surface of the textile to be inspected and the global structure encoding features of the surface of the textile to be inspected to obtain multi-scale image encoding features of the surface of the textile to be inspected; The alarm module is used to obtain the detection result and determine whether to issue an alarm signal based on the multi-scale image coding features of the surface of the textile to be detected.

2. The textile state detection and analysis system according to claim 1, characterized in that: The textile surface image enhancement module is used to: perform noise suppression and illumination unevenness correction on the textile surface image to be inspected to obtain the enhanced textile surface image to be inspected.

3. The textile state detection and analysis system according to claim 1, characterized in that: The multi-scale feature extraction unit is used to: use a multi-scale feature extractor based on a dilated spatial pyramid pooling model to respectively extract local detail information and global abstract information from the enhanced surface image of the textile to be inspected to obtain a local detail coding feature map of the surface of the textile to be inspected as the local detail coding feature of the surface of the textile to be inspected and a global structure coding feature map of the surface of the textile to be inspected as the global structure coding feature of the surface of the textile to be inspected.

4. The textile state detection and analysis system according to claim 1, characterized in that: The multi-scale feature fusion unit comprises: The feature decoupling subunit of the textile to be inspected is used to perform feature decoupling on the local detail coding feature map of the surface of the textile to be inspected and the global structure coding feature map of the surface of the textile to be inspected to obtain a set of local detail feature coding matrices of the surface of the textile to be inspected and a set of local feature coding matrices of the global structure of the surface of the textile to be inspected; The multi-scale image feature alignment and fusion subunit is used to perform phase alignment and attention saliency fusion on the set of local detail feature coding matrices of the surface of the textile to be inspected and the set of global structure local feature coding matrices of the surface of the textile to be inspected to obtain a multi-scale image coding feature map of the surface of the textile to be inspected as the multi-scale image coding feature of the surface of the textile to be inspected.

5. The textile state detection and analysis system according to claim 4, characterized in that: The multi-scale image feature alignment and fusion subunit includes: A feature phase dynamic search and alignment secondary subunit is used to perform feature phase dynamic search and alignment on the set of local detail feature coding matrices of the textile surface to be inspected and the set of local feature coding matrices of the global structure of the textile surface to be inspected based on the feature phase alignment between any two local detail feature coding matrices of the textile surface to be inspected and the local feature coding matrices of the global structure of the textile surface to be inspected in the set of local detail feature coding matrices of the textile surface to be inspected and the set of local feature coding matrices of the global structure of the textile surface to be inspected, so as to obtain a set of phase-aligned {local detail feature coding matrices of the textile surface to be inspected, local feature coding matrices of the global structure of the textile surface to be inspected} feature pairs; The attention weight calculation secondary subunit is used to input each phase-aligned {local detail feature coding matrix of the surface of the textile to be inspected, local feature coding matrix of the global structure of the surface of the textile to be inspected} feature pair in the set of phase-aligned {local detail feature coding matrix of the surface of the textile to be inspected, local feature coding matrix of the global structure of the surface of the textile to be inspected} feature pairs into a feature joint perception attention network to obtain a set of joint perception attention weights of the surface features of the textile to be inspected; The secondary subunit of multi-scale salient aggregation of the surface of the textile to be inspected is used to perform attention-driven salient aggregation on the set of phase-aligned {local detail feature coding matrix of the surface of the textile to be inspected, local feature coding matrix of the surface of the textile to be inspected} feature pairs based on the set of joint perceptual attention weights of the surface features of the textile to be inspected, so as to obtain a multi-scale image coding feature map of the surface of the textile to be inspected.

6. The textile state detection and analysis system according to claim 5, characterized in that: The attention weight calculation secondary subunit includes: The third-level subunit for calculating scale association of the surface of the textile to be inspected is used to calculate the association matrix between each phase-aligned {local detail feature coding matrix of the surface of the textile to be inspected, local feature coding matrix of the global structure of the surface of the textile to be inspected} feature pair in the set of phase-aligned {local detail feature coding matrix of the surface of the textile to be inspected, local feature coding matrix of the global structure of the surface of the textile to be inspected} feature pairs to obtain a set of scale association matrices of the surface of the textile to be inspected; The third-level subunit of the surface trace measurement of the textile to be inspected is used to perform trace measurement on each of the surface pair scale association matrices of the textile to be inspected in the set of the surface pair scale association matrices of the textile to be inspected to obtain a set of surface trace measurement values ​​of the textile to be inspected; The third-level subunit for normalizing the trace measurement weights is used to obtain a set of joint perceptual attention weights of the surface features of the textile to be inspected based on the set of trace measurement values ​​of the surface of the textile to be inspected.

7. The textile state detection and analysis system according to claim 6, characterized in that: The trace value weight normalization three-level subunit is used to: Based on the set of surface trace measurement values ​​of the textile to be inspected, each phase-aligned {local detail feature coding matrix of the textile to be inspected surface, local feature coding matrix of the global structure of the textile to be inspected surface} feature pair in the set of phase-aligned {local detail feature coding matrix of the textile to be inspected surface, local feature coding matrix of the global structure of the textile to be inspected surface} feature pairs is subjected to matrix manifold optimization based on the number of dual connections to obtain a set of phase-optimized aligned {local detail feature coding matrix of the textile to be inspected surface, local feature coding matrix of the global structure of the textile to be inspected surface} feature pairs; Based on the phase optimization alignment of the set of feature pairs {local detail feature encoding matrix of the surface of the textile to be inspected, local feature encoding matrix of the global structure of the surface of the textile to be inspected}, a set of joint perceptual attention weights of the surface features of the textile to be inspected is obtained.

8. The textile state detection and analysis system according to claim 7, characterized in that: The alarm module is used for: Inputting the multi-scale image coding feature map of the surface of the textile to be inspected into the trained defect detection model to obtain a detection result; Based on the detection result, it is determined whether to issue an alarm signal.

9. The textile state detection and analysis system according to claim 8, characterized in that: The detection result includes the defect point location, defect point type and confidence level.

Images