Fabric defect intelligent detection method and system based on AI visual identification

Abstract

The invention relates to the technical field of fabric detection, and discloses a fabric defect intelligent detection method and system based on AI visual identification. According to the method, motion blur is quantized through motion state data, optical blur caused by fabric motion is eliminated through deconvolution solution, so that motion interference in the fabric transmission process is processed in a targeted mode, self-adaptive balance of the deblurring capacity and the feature retention capacity is achieved, and then based on the optical interference principle, the deblurring capacity and the feature retention capacity are improved. Through a dynamic calibration system combining hardware-level real-time compensation and multi-dimensional optical parameter calibration, dynamic optical parameter calibration of primary correction data is realized, then fabric defect characterization data is extracted to accurately obtain defect features, and finally, a detection-production line control closed loop is constructed through a quality quantitative index and a comprehensive risk value, so that fabric defect detection is realized. The fabric defect detection precision can be improved, so that the problem of high defect missing detection and false detection rate caused by optical data distortion due to movement and environment interference in a traditional method is effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of fabric detection, and in particular to an intelligent fabric defect detection method and system based on AI visual recognition. Background Art

[0002] Fabric defects refer to defects that affect the appearance and performance of fabrics caused by process deviations, equipment failures or environmental factors during the production and processing (such as spinning, weaving, dyeing and finishing), and common types include yarn breakage, oil stains, holes, uneven yarn density, stains, etc.; intelligent detection of fabric defects based on AI visual recognition can replace traditional manual inspection, which can not only solve the problems of low efficiency and inconsistent inspection standards of manual inspection, but also reduce the production costs of enterprises and ensure the stability of product quality; intelligent detection of fabric defects based on AI visual recognition refers to the use of high-resolution industrial cameras with specific light sources to collect surface images of moving fabrics, and then use image preprocessing algorithms and deep learning models to detect and classify defects, so as to achieve precise control of fabric quality.

[0003] In traditional fabric defect detection, changes in the fabric's motion state during transmission and environmental interference (such as mechanical vibration and light fluctuations) will destroy the optical characteristics of fabric defects under specific wavelengths of visible light (such as differences in reflectivity and absorption), resulting in distortion of the optical contrast between defects and normal areas in the optical response data, leading to a high rate of missed and false detection of defects. Summary of the Invention

[0004] The main purpose of this invention is to provide an intelligent fabric defect detection method and system based on AI visual recognition, aiming to solve the technical problems in the prior art.

[0005] The present invention proposes an intelligent fabric defect detection method based on AI visual recognition, comprising: Based on AI visual recognition, it uses visible light of a specific wavelength to obtain optical response data on the fabric surface, and uses sensors to obtain fabric motion state data and environmental interference data; acquiring primary correction data according to the motion state data and the optical response data; Performing dynamic optical parameter calibration on the primary correction data according to the environmental interference data to obtain optical feature enhancement data; Acquiring fabric defect characterization data based on the optical feature enhancement data, wherein the fabric defect characterization data includes a high-frequency texture feature map and a defect comprehensive characterization map; Obtaining a quality quantification index based on the high-frequency texture feature map and the defect comprehensive characterization map, and obtaining a comprehensive risk value based on the quality quantification index; Compare the comprehensive risk value with the preset threshold: If the comprehensive risk value is not greater than a preset threshold, outputting the quality quantitative index; If the comprehensive risk value is greater than a preset threshold, graded response control information is generated according to the quality quantification index, and adjustment operations are performed on the textile production line according to the graded response control information, wherein the graded response control information includes a speed reduction ratio and a light source adjustment parameter.

[0006] Preferably, the step of acquiring primary correction data according to the motion state data and the optical response data comprises: Acquire a motion speed parameter, a motion direction, and a motion period of the fabric according to the motion state data, and acquire a motion blur length of the fabric according to the motion speed parameter; Obtaining an optical detection distance and a static blur base value, and obtaining an optical blur characteristic coefficient according to the optical detection distance, the static blur base value and the motion blur length; Obtaining a time-varying blur kernel matrix according to the optical blur characteristic coefficient and the fabric movement direction; Acquire a sequence of overlapping subframes according to the optical response data and the fabric motion period; Obtaining a vibration sensitivity coefficient, and obtaining a dynamic noise parameter according to the vibration sensitivity coefficient and the motion speed parameter; Acquire a defect characteristic response spectrum, and acquire a denoised subframe sequence according to the defect characteristic response spectrum, the overlapping subframe sequence and the dynamic noise parameter; A deblurred subframe sequence is obtained according to the motion speed parameter, the time-varying blur kernel matrix and the denoising subframe sequence, and primary correction data is obtained according to the fabric motion period and the deblurred subframe sequence.

[0007] Preferably, the step of performing dynamic optical parameter calibration on the primary correction data according to the environmental interference data to obtain optical feature enhancement data comprises: Acquiring mechanical vibration parameters and ambient light parameters according to the environmental interference data; Obtaining a calibration conversion coefficient, and obtaining a reflectivity fluctuation prediction value based on the calibration conversion coefficient and the mechanical vibration parameter; Acquiring optical detection reference parameters, and performing phase-sensitive demodulation processing on the primary correction data according to the optical detection reference parameters and the reflectivity fluctuation prediction value to obtain an anti-disturbance reflectivity distribution matrix; Obtaining a dynamic light disturbance amount and a spectral shift amount according to the ambient light parameter, and obtaining a dynamic compensation coefficient and a wavelength compensation coefficient according to the dynamic light disturbance amount, the spectral shift amount and the optical detection reference parameter; Optical feature fusion is performed on the primary correction data according to the anti-disturbance reflectivity distribution matrix, the dynamic compensation coefficient and the wavelength compensation coefficient to obtain optical feature enhancement data.

[0008] Preferably, the step of obtaining fabric defect characterization data based on the optical feature enhancement data comprises: Obtaining the RGB value of each pixel according to the optical feature enhancement data; Obtaining an optical feature-noise correlation database and fabric benchmark spectra; Obtaining a minimum detection size of a defect, and obtaining an optical characteristic component based on the RGB value corresponding to each pixel, a fabric reference spectrum, and the minimum detection size of the defect; Acquire an optimized characteristic component according to the optical characteristic-noise correlation database and the optical characteristic component, wherein the optimized characteristic component includes a substrate reflection component, an abnormal absorption component, and a high-frequency scattered noise component; A high-frequency texture feature map is obtained based on the fabric reference spectrum and the abnormal absorption component, and a comprehensive defect characterization map is obtained based on the base reflection component, the abnormal absorption component and the high-frequency scattering noise component.

[0009] Preferably, the step of obtaining a quality quantification index based on the high-frequency texture feature map and the defect comprehensive characterization map, and obtaining a comprehensive risk value based on the quality quantification index includes: Obtaining an optical signal difference value for each pixel according to the high-frequency texture feature map, and obtaining a defect heat map according to the optical signal difference value, the attention weight matrix, and the defect comprehensive representation map; Obtaining defect optical characterization data according to the defect thermal map, and obtaining a defect detection result table according to the defect optical characterization data; Using the defect heat map and the defect detection result table as quality quantitative indicators; Obtaining a defect density distribution matrix according to the defect heat map, and obtaining defect size, defect type, and defect coordinates according to the defect density distribution matrix and the defect comprehensive characterization map; A defect impact index is obtained according to the defect size, defect type and defect coordinates, and a comprehensive risk value is obtained according to the defect density distribution matrix, defect type and the defect impact index.

[0010] Preferably, the step of generating hierarchical response control information according to the quality quantification indicator includes: Obtaining defect characteristic energy and defect diffusion rate according to the quality quantification index, and obtaining high-frequency noise energy according to the environmental interference data; Obtaining optical characteristic stability according to the high-frequency noise energy and the defect characteristic energy, and determining whether the optical characteristic stability is greater than a stability threshold; If the optical characteristic stability is greater than the stability threshold, obtaining the optical characteristic parameters of the defect according to the defect comprehensive characterization map; Obtaining a fabric optical property reference library, and obtaining a reflectivity anomaly amplitude, a transmittance gradient change, and an absorption peak offset based on the fabric optical property reference library and the optical characteristic parameters; Obtaining an optical sensitivity level according to the reflectivity anomaly amplitude, the transmittance gradient change, and the absorption peak shift; A graded response database is obtained, and graded response control information is obtained according to the graded response database, the optical sensitivity level, and the defect diffusion rate.

[0011] This application also provides an intelligent fabric defect detection system based on AI visual recognition, including: The data acquisition module is used to obtain optical response data of the fabric surface using visible light of a specific wavelength based on AI visual recognition, and obtain the fabric's motion state data and environmental interference data through sensors; a motion correction module, configured to obtain primary correction data based on the motion state data and the optical response data; a parameter calibration module, configured to perform dynamic optical parameter calibration on the primary correction data according to the environmental interference data to obtain optical feature enhancement data; a defect analysis module, configured to obtain fabric defect characterization data based on the optical feature enhancement data, wherein the fabric defect characterization data includes a high-frequency texture feature map and a defect comprehensive characterization map; a detection and evaluation module, configured to obtain a quality quantitative index based on the high-frequency texture feature map and the defect comprehensive characterization map, and obtain a comprehensive risk value based on the quality quantitative index; A detection and control module is used to determine whether the comprehensive risk value is greater than a preset threshold; If the comprehensive risk value is not greater than a preset threshold, outputting the quality quantitative index; If the comprehensive risk value is greater than a preset threshold, graded response control information is generated according to the quality quantification index, and adjustment operations are performed on the textile production line according to the graded response control information, wherein the graded response control information includes a speed reduction ratio and a light source adjustment parameter.

[0012] Preferably, the motion correction module comprises: A motion sensing unit, configured to obtain a motion speed parameter, a motion direction, and a motion period of the fabric according to the motion state data, and to obtain a motion blur length of the fabric according to the motion speed parameter; a fuzzy quantization unit, configured to obtain an optical detection distance and a static fuzzy base value, and obtain an optical fuzzy characteristic coefficient according to the optical detection distance, the static fuzzy base value, and the motion blur length; A matrix construction unit, configured to obtain a time-varying blur kernel matrix according to the optical blur characteristic coefficient and the fabric movement direction; a subframe interception unit, configured to obtain an overlapping subframe sequence according to the optical response data and the fabric motion period; a noise evaluation unit, configured to obtain a vibration sensitivity coefficient and obtain a dynamic noise parameter according to the vibration sensitivity coefficient and the motion speed parameter; a noise suppression unit, configured to obtain a defect characteristic response spectrum, and obtain a denoised subframe sequence according to the defect characteristic response spectrum, the overlapping subframe sequence and the dynamic noise parameter; A data correction unit is configured to obtain a deblurred subframe sequence according to the motion speed parameter, the time-varying blur kernel matrix and the denoised subframe sequence, and to obtain primary correction data according to the fabric motion period and the deblurred subframe sequence.

[0013] The present invention also provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned intelligent detection method for fabric defects based on AI visual recognition are implemented.

[0014] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned method for intelligent detection of fabric defects based on AI visual recognition.

[0015] The beneficial effects of the present invention are as follows: the present invention obtains the optical response data of the fabric surface, the motion state data of the fabric and the environmental interference data, thereby providing a comprehensive data basis for the subsequent elimination of motion and environmental interference and restoration of the true optical characteristics of the defect; then, primary correction data is obtained based on the motion state data and the optical response data, and the motion interference in the fabric transmission process is targeted to preliminarily restore the true optical characteristics of the defect, thereby solving the problem in the prior art that "fabric movement causes optical blur, subframe missed detection, and distorts the optical contrast between the defect and the normal area"; then, dynamic optical parameter calibration is performed on the primary correction data based on the environmental interference data to solve the problem in the prior art that "environmental interference destroys the optical characteristics of the defect, resulting in the optical contrast between the defect and the normal area in the optical response data is distorted". The problem of "distortion" is solved. Then, fabric defect characterization data including high-frequency texture feature maps and defect comprehensive characterization maps are obtained based on the optical feature enhancement data, which realizes the accurate extraction of defect features. Based on this, quality quantitative indicators and comprehensive risk values ​​can be further obtained, thereby providing a quantifiable basis for subsequent quality judgment and production line adjustment. Finally, by judging the relationship between the comprehensive risk value and the preset threshold, quality quantitative indicators are output or graded response control information is generated and executed to build a closed loop from detection to production line control, realizing the linkage between detection and production control. Through the synergistic effect of the above contents, the accuracy and anti-environmental interference ability of fabric defect detection are comprehensively improved, and the problem of optical data distortion caused by motion and environmental interference in traditional detection methods, resulting in a high rate of missed detection and false detection of defects, is effectively solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Schematic diagram of a method flow according to an embodiment of the present invention.

[0017] Figure 2 FIG. 1 is a schematic diagram of a system structure according to an embodiment of the present invention.

[0018] Figure 3 This is a schematic diagram of the internal structure of a computer device according to an embodiment of the present application.

[0019] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0020] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0021] like Figure 1 As shown, this application provides an intelligent fabric defect detection method based on AI visual recognition, including: S1. Based on AI visual recognition, it uses visible light of a specific wavelength to obtain optical response data of the fabric surface, and obtains the fabric's motion state data and environmental interference data through sensors; S2. acquiring primary correction data according to the motion state data and the optical response data; S3. Performing dynamic optical parameter calibration on the primary correction data according to the environmental interference data to obtain optical feature enhancement data; S4. Acquiring fabric defect characterization data based on the optical feature enhancement data, wherein the fabric defect characterization data includes a high-frequency texture feature map and a defect comprehensive characterization map; S5. Obtaining a quality quantification index based on the high-frequency texture feature map and the defect comprehensive characterization map, and obtaining a comprehensive risk value based on the quality quantification index; S6. Compare the comprehensive risk value with a preset threshold: If the comprehensive risk value is not greater than a preset threshold, outputting the quality quantitative index; If the comprehensive risk value is greater than a preset threshold, graded response control information is generated according to the quality quantification index, and adjustment operations are performed on the textile production line according to the graded response control information, wherein the graded response control information includes a speed reduction ratio and a light source adjustment parameter.

[0022] As described in the above steps S1-S6, the present invention determines a specific wavelength of visible light according to the fabric material (such as blue light is preferred for cotton and polyester, which can maximize the difference in reflectance / absorption rate between defects (such as oil stains and broken yarns) and normal areas), and uses a specific wavelength of visible light (such as blue light) adapted to the fabric material to irradiate the fabric surface, and obtains optical response data of the fabric surface based on AI visual recognition, wherein the optical response data refers to multiple frames of optical images of the fabric surface continuously collected under the irradiation of a single specific wavelength (such as blue light, near-infrared light, etc.), so as to utilize the difference in reflectance / absorption rate between defects and normal areas at the wavelength to form original detection data, and simultaneously obtains the motion state data and environmental interference data of the fabric in combination with the sensor data of the textile production line, wherein , motion state data refers to the parameter set collected by sensors and used to quantify the dynamic characteristics of fabrics during transmission. Environmental interference data refers to the external environmental parameter set collected by sensors that may destroy the optical characteristics of fabric defects, mainly including mechanical vibration parameters and ambient light parameters. By obtaining multi-dimensional data to break the limitation of traditional detection relying only on single optical data, it is helpful to solve the problem of "no correlation between motion state and environmental interference data, and inability to trace the cause of optical response distortion" in existing technologies, and provides a comprehensive data foundation for the subsequent elimination of motion and environmental interference and restoration of the true optical characteristics of defects. Then, for the motion interference in the fabric transmission process, primary correction data is obtained based on the motion state data and optical response data. Among them, the initial The primary correction data refers to the image data that preliminarily restores the real optical characteristics of the fabric after motion data correction. Through this correction of motion blur and sub-frame optimization, the real optical characteristics of the defects are preliminarily restored, thereby solving the problem in the prior art that "fabric movement causes optical blur, sub-frame missed detection, and distorts the optical contrast between defects and normal areas". Then, in order to consider the influence of environmental interference on optical detection, the primary correction data is dynamically optically calibrated according to the environmental interference data to obtain optical feature enhancement data. Among them, the optical feature enhancement data refers to the image data obtained after dynamic optical parameter calibration of the primary correction data according to the environmental interference data, which further restores the real optical characteristics of the fabric. This method strengthens the optical characteristics of defects through dynamic calibration. The method is conducive to solving the problem of "environmental interference destroying the optical characteristics of defects, resulting in distortion of the optical contrast between defects and normal areas in the optical response data" in the existing technology. Then, based on the optical feature enhancement data, fabric defect characterization data including high-frequency texture feature maps and defect comprehensive characterization maps are obtained, which realizes the accurate extraction of defect features, thereby providing a clear and reliable feature basis for defect detection. Among them, fabric defect characterization data refers to structured data extracted from optical feature enhancement data for describing fabric defect characteristics, and high-frequency texture feature maps refer to images used to highlight subtle structural changes on the fabric surface. Its core focuses on the high-frequency signal differences generated by defects (such as yarn breakage, stains, holes, etc.) in the normal texture of the fabric.The defect comprehensive characterization map refers to a map that comprehensively reflects the authenticity and severity of defects. Since the existing technology has the problem of "fuzzy assessment of the impact of defects on quality and lack of quantitative indicators", the quality quantitative indicators are subsequently obtained based on the high-frequency texture feature map and the defect comprehensive characterization map. Among them, the quality quantitative indicators refer to specific data used to quantify the quality status of fabric defects, including defect heat maps and defect detection result tables, and further obtain comprehensive risk values ​​to quantitatively evaluate the impact of defects on quality, and provide a quantifiable basis for subsequent quality judgment and production line adjustments. Finally, by judging the relationship between the comprehensive risk value and the preset threshold, the quality quantitative indicators are output or the classification is generated and executed. Response control information, where hierarchical response control information refers to a set of instructions used to guide adjustments to textile production lines, including speed reduction ratios and light source adjustment parameters. The speed reduction ratio refers to the percentage parameter used to adjust the fabric transmission speed, and the light source adjustment parameters refer to the parameters used to optimize the detection light source, including light intensity and characteristic wavelength. By using hierarchical response control information to build a closed loop from detection to production line control, the linkage between detection and production control is achieved. Through the synergistic effect of the above contents, the accuracy of fabric defect detection and its ability to resist environmental interference are comprehensively improved, effectively solving the problem of optical data distortion caused by motion and environmental interference in traditional detection methods, resulting in high defect missed detection and false detection rates.

[0023] In one embodiment, the step S2 of acquiring primary correction data according to the motion state data and the optical response data includes: S21, obtaining a motion speed parameter, a motion direction, and a motion period of the fabric according to the motion state data, and obtaining a motion blur length of the fabric according to the motion speed parameter; S22, obtaining an optical detection distance and a static blur base value, and obtaining an optical blur characteristic coefficient according to the optical detection distance, the static blur base value, and the motion blur length; S23, obtaining a time-varying blur kernel matrix according to the optical blur characteristic coefficient and the fabric movement direction; S24, acquiring an overlapping subframe sequence according to the optical response data and the fabric motion period; S25. Obtain a vibration sensitivity coefficient, and obtain a dynamic noise parameter according to the vibration sensitivity coefficient and the motion speed parameter; S26. Acquire a defect characteristic response spectrum, and acquire a denoised subframe sequence according to the defect characteristic response spectrum, the overlapping subframe sequence, and the dynamic noise parameter; S27. Acquire a deblurred subframe sequence according to the motion speed parameter, the time-varying blur kernel matrix, and the denoising subframe sequence, and acquire primary correction data according to the fabric motion period and the deblurred subframe sequence.

[0024] As described in the above steps S21-S27, the present invention obtains the motion speed parameters of the fabric through the motion state data, wherein the motion speed parameters refer to physical parameters that can quantitatively describe the speed and change trend of the position of the fabric during the motion process, including real-time linear velocity and acceleration, and then obtains the optical detection distance and static blur base value, wherein the optical detection distance refers to the vertical distance from the camera lens to the fabric surface in the optical system, which directly affects the imaging magnification and depth of field, and the static blur base value refers to the inherent blur degree of the optical system in a static state determined by the calibration experiment. Then, based on the kinematic principles and optical imaging model, the real-time linear velocity and acceleration are converted into blur length, wherein the blur length refers to the physical length of the shadow of a single pixel in the optical image caused by the motion of the fabric, and its calculation formula is: ,in, Indicates the fuzzy length, Indicates the real-time line speed, represents the square of the camera exposure time, Indicates acceleration, the “ " indicates the linear blur caused by uniform motion, which is the dominant term of blur length. When the acceleration of the fabric movement is large (such as sudden start and stop), additional compensation for blur length is required, so " " represents the nonlinear blur compensation term caused by acceleration. Then, the ratio of blur length to optical detection distance is calculated, and the sum of this ratio and static blur base value is calculated to obtain the optical blur characteristic coefficient. The optical blur characteristic coefficient refers to the standardized blur length that characterizes the intensity and scale of blur. The optical blur characteristic coefficient is used to eliminate the geometric differences of the imaging system (such as the blur scale changes caused by different object distances) and static noise interference. The above method realizes the accurate quantification of fabric imaging blur under complex motion state, thus providing a unified and accurate quantitative basis for subsequent blur correction. Furthermore, the fabric movement direction is obtained according to the motion state data, and a discretized time-varying blur kernel matrix is ​​generated according to the fabric movement direction and the optical blur characteristic coefficient. The dimension of the matrix is ,in, Represents the optical blur characteristic coefficient. Its non-zero elements are arranged in a single-pixel width line segment uniformly distributed along the direction of motion to simulate the pixel smear effect caused by motion. The sum of the matrix element values ​​is 1 to ensure energy conservation. For example, when moving horizontally to the right (the direction of fabric movement is equal to 0°) and the optical blur characteristic coefficient is equal to 5, the first 5 elements of the center row of the matrix are 1 / 5, and the rest are 0. The core content of the time-varying blur kernel matrix includes the motion direction parameter (determines the geometric direction of the blur smear), the blur intensity parameter (controls the physical scale of the blur kernel) and the time-varying update mechanism (adjusts the optical blur characteristic coefficient and The time-domain update frequency of the time-varying blur kernel matrix is ​​dynamically adjusted through real-time acceleration feedback (e.g., high-frequency updates to cope with sudden accelerations), ensuring that the blur kernel always accurately represents the current motion state. The time-varying blur kernel matrix provides an accurate degradation model for the subsequent deconvolution deblurring algorithm. The convolution operation also reduces optical feature distortion caused by fabric motion (e.g., smear at yarn breaks), significantly improving the accuracy of defect detection. This addresses the problem of traditional technologies being unable to adapt to dynamic changes in fabric motion direction and speed, resulting in unstable blur correction effects. The fabric motion cycle (i.e., the minimum time unit of the fabric's repetitive motion) is then acquired based on the motion state data. The optical response data is then segmented into several time segments based on the motion cycle. Within each segment, a continuous subframe sequence is captured at a preset overlap rate (e.g., 30% to 50%) to ensure that adjacent subframes overlap to cover the entire fabric surface. This results in an overlapping subframe sequence. An overlapping subframe sequence is a collection of partially overlapping images captured from continuous images based on the fabric motion cycle and displacement compensation. The following conditions must be met during capture: the subframe size must be larger than the maximum physical size of the defect, and the exposure parameters must be adapted to a specific wavelength (e.g., blue light enhances the contrast of oil stains) to preserve the reflectivity / transmittance difference between the defect and normal areas. If the motion cycle changes due to variable-speed production, the subframe capture frequency and overlap rate are recalculated to ensure there are no blind spots. The resulting overlapping subframe sequence serves as input for subsequent deblurring and feature extraction, providing a high-fidelity optical data foundation for defect identification. Then, based on the instantaneous acceleration of the fabric during movement, the signal-to-noise ratio of each subframe is calculated synchronously (determined by the ratio of the variance to the mean of the grayscale value of the subframe image), and combined with the pre-acquired vibration sensitivity coefficient (the grayscale fluctuation caused by unit acceleration measured experimentally) and the maximum acceleration of the system (the upper limit of mechanical vibration allowed by the hardware), the instantaneous acceleration is mapped to the vibration noise intensity component through normalization processing, and the inverse of the subframe signal-to-noise ratio is further introduced to characterize the inherent noise of the optical system (such as light fluctuation and sensor noise). Finally, the dynamic noise parameter is calculated according to the vibration noise intensity component and the inverse of the subframe signal-to-noise ratio through the weighted fusion formula (the weight coefficients corresponding to the vibration noise intensity component and the inverse of the subframe signal-to-noise ratio are determined by the calibration experiment). Among them, the dynamic noise Acoustic parameters refer to quantitative indicators that combine vibration noise and optical noise. Dynamic noise parameters will be used for the subsequent adaptive adjustment of the defect characteristic response spectrum to ensure that noise interference can still be effectively suppressed under variable speed or high vibration conditions, maintaining the stability of the defect optical characteristics. Traditional methods cannot distinguish between different types of noise. Under variable speed or high vibration conditions, defect characteristics are easily misjudged as noise filtering, or due to insufficient noise reduction, defect characteristics are masked by noise, seriously affecting detection stability. By constructing dynamic noise parameters, this solution achieves comprehensive quantification and differentiation of vibration noise and inherent noise of the optical system. By associating dynamic noise parameters with the subsequent filtering process, noise suppression can be adaptively adjusted according to the real-time noise status, effectively reducing noise while maximally retaining the defect optical characteristics. Subsequently, an optical system calibration experiment is conducted to obtain a defect characteristic response spectrum adapted to the target wavelength (such as 450nm blue light). The defect characteristic response spectrum refers to the ability of the optical system to transmit light intensity distributions of different spatial frequencies, and includes two parts: modulus (modulation transfer function) and phase. It also characterizes the transmission efficiency of the system to defect characteristics (such as sudden changes in oil absorption rate and differences in broken yarn transmittance) in the form of frequency domain. Each frame image in the overlapping subframe sequence is then converted to the frequency domain through Fourier transform, and the image spectrum is weighted filtered using the modulus curve (MTF) of the defect characteristic response spectrum to enhance the mid- and high-frequency components corresponding to the defect characteristics (such as the 20-50LP / mm frequency band), while suppressing the high-frequency noise band identified by the dynamic noise parameters. For periodic noise caused by vibration (such as the main frequency of mechanical vibration), a notch filter is designed in the frequency domain, and local spectrum repair is performed with reference to the phase information of the defect characteristic response spectrum. Finally, The spatial domain image is reconstructed through inverse Fourier transform to generate a denoised subframe sequence. The denoised subframe sequence refers to a set of subframes that have been processed through frequency domain filtering of the defect feature response spectrum and dynamic noise suppression. In this process, the dynamic noise parameter is used to adaptively adjust the filtering threshold: when the dynamic noise parameter is high, the low-pass characteristics of the defect feature response spectrum are enhanced to suppress vibration noise. Conversely, the frequency band is widened to retain subtle defect features, thereby maximizing the optical contrast between the defect and the background while denoising. Traditional spatial domain filtering has difficulty in effectively separating defect features from noise, especially in strong interference environments such as vibration. The optical contrast between the defect and the background is severely damaged, making defect identification difficult and unable to selectively enhance defect features at specific wavelengths. However, this solution uses a method of adaptively adjusting the filtering threshold in combination with dynamic noise parameters to balance noise reduction and defect feature preservation, which is conducive to further improving the denoising effect. Then, the basic weight is obtained. The basic weight refers to the optical reflectance benchmark value of the normal fabric texture area under a specific wavelength, and is normalized and calibrated in combination with the system imaging signal-to-noise ratio. Subsequently, the ratio of the real-time linear speed to the maximum design speed of the system is introduced as a dynamic adjustment factor, and the exponential decay function is used to calculate the basic weight. ” Calculate the final regularization weights, where Represents the regularization weight, which is a key hyperparameter for balancing the algorithm's "defuzzification ability" and "feature retention ability". represents the basic weight, represents the natural exponential function, Indicates the attenuation coefficient, usually 1.5~2, Indicates the real-time line speed, The physical meaning of this formula is that when the real-time linear speed of the fabric approaches the maximum design speed of the system, the regularization weight is automatically reduced to enhance the aggressiveness of the deblurring algorithm (prioritizing the elimination of motion smear), while at low speeds the regularization weight is increased to protect optical texture details (such as the micron-level transmittance features at yarn breaks). The regularization weight will directly act on the subsequent deconvolution optimization objective function to achieve an adaptive balance between motion blur correction and feature retention, ensuring that the optical characteristic parameters of defects can still be stably extracted under variable speed production environments (such as the absorption rate deviation of the stain area in the blue light band ≥ 25%). Finally, based on the denoised sub-frame sequence, the time-varying blur kernel matrix, and the regularization weight, an optimization objective function is constructed with the goal of minimizing the defect feature restoration error. The deconvolution solution is performed by combining the initial estimated image (a reference image generated based on the optical reflectance / transmittance modeling of normal fabrics at specific wavelengths, which is used to provide physical constraints for deconvolution and avoid algorithm divergence or artifact generation) through the Richardson-Lucy iterative algorithm. The deblurred sub-frame sequence is gradually optimized. This method can eliminate the defects caused by fabric smearing. The optical blur caused by motion (such as smear and texture diffusion) is eliminated, and the difference in optical properties (such as reflectivity and transmittance contrast) between defect areas (such as holes and oil stains) and normal areas is significantly enhanced to ensure that the physical value error of defect features (such as areas with increased transmittance of holes and areas with reduced absorption of oil stains) is ≤5%. Subsequently, the ratio of the fabric motion period to the number of sub-frames is calculated to obtain the sub-frame time interval, and the product of the sub-frame time interval and the real-time linear velocity is calculated to obtain the displacement vector. The phase correlation method is then used to perform sub-pixel registration on the deblurred sub-frame sequence to eliminate mechanical vibration. The misalignment caused by the movement is finally generated through a weighted stitching algorithm (the weight is assigned by the sub-frame signal-to-noise ratio), and the primary correction data must meet the following requirements: a single defect (such as a hole with a diameter ≥ 0.5 mm) has no break or repetition in the image, and the relative difference in reflectivity / transmittance between the defect area and the normal area is retained at ≥ 95%. Through this regularization weight dynamic adjustment mechanism based on real-time linear speed, an adaptive balance between deblurring capability and feature retention capability is achieved, the deblurring accuracy is improved, and artifact generation is avoided, thereby ensuring the integrity and accuracy of the primary correction data.

[0025] In one embodiment, the step S3 of performing dynamic optical parameter calibration on the primary correction data according to the environmental interference data to obtain optical feature enhancement data includes: S31, obtaining mechanical vibration parameters and ambient light parameters according to the environmental interference data; S32, obtaining a calibration conversion coefficient, and obtaining a reflectivity fluctuation prediction value according to the calibration conversion coefficient and the mechanical vibration parameter; S33, obtaining optical detection reference parameters, and performing phase-sensitive demodulation processing on the primary correction data according to the optical detection reference parameters and the reflectivity fluctuation prediction value to obtain an anti-disturbance reflectivity distribution matrix; S34, obtaining a dynamic light disturbance amount and a spectral shift amount according to the ambient light parameter, and obtaining a dynamic compensation coefficient and a wavelength compensation coefficient according to the dynamic light disturbance amount, the spectral shift amount and the optical detection reference parameter; S35. Perform optical feature fusion on the primary correction data according to the anti-disturbance reflectivity distribution matrix, the dynamic compensation coefficient, and the wavelength compensation coefficient to obtain optical feature enhancement data.

[0026] As described in the above steps S31-S35, the present invention obtains mechanical vibration parameters and ambient light parameters through environmental interference data, wherein mechanical vibration parameters refer to physical parameters that can quantitatively describe the mechanical vibration properties that affect the stability of the optical detection system, including the main vibration frequency, vibration amplitude and vibration direction angle; ambient light parameters refer to physical parameters that can quantitatively describe the light properties that interfere with the optical detection accuracy caused by changes in ambient light, including dynamic light disturbance, spectral offset and flicker frequency; then a calibration conversion coefficient is obtained through a laser Doppler vibrometer calibration system, wherein the calibration conversion coefficient characterizes the optical path change corresponding to the unit output voltage of the vibration sensor, and the main vibration frequency, vibration amplitude and vibration direction angle are input into the FPGA processing unit, combined with the calibration conversion coefficient, and the optical path difference change caused by the vibration is calculated according to the principle of optical interference, wherein the optical path difference change refers to the periodic change in the optical path length caused by the vibration, which is used to quantify the time-varying interference intensity of the vibration on the optical detection system and provide input parameters for the reflectivity modulation model, and then Then, a standard reflector is used to calibrate the fabric under static conditions to obtain a baseline reflectivity distribution. Based on the data obtained above, a reflectivity modulation model is established based on Fresnel reflection theory. The reflectivity modulation model refers to a mathematical model that describes the quantitative relationship between the change in optical path difference caused by vibration and the fluctuation in fabric reflectivity. The model linearly maps the optical path difference to reflectivity fluctuations through Taylor's first-order approximation and outputs a predicted value of reflectivity fluctuations caused by vibration in real time. The predicted value of reflectivity fluctuations refers to quantitative data calculated by the model that reflects the change in reflectivity of each point on the fabric over time due to vibration, including fluctuation amplitude, frequency and phase information. Verified by Polytec OFV-505, the phase deviation between the model prediction value and the measured data is ≤0.1π, so as to provide a high-precision interference feature template for optical compensation to solve the problem that traditional methods cannot quantify the impact of mechanical vibration on reflectivity measurement, resulting in reflectivity fluctuations caused by vibration being misjudged as fabric defect characteristics, or the real defect characteristics being masked by vibration noise, seriously affecting the detection accuracy. Then, a scientific-grade CCD camera (such as Hamamatsu Orca-Flash4.0) is used to obtain the original voltage matrix based on the primary calibration data. The original voltage matrix refers to a two-dimensional array composed of digitized voltage values ​​output by photoelectric conversion, including the original voltage values ​​corresponding to multiple pixel points. The dark field reference voltage matrix, reference voltage matrix and calibrated reflectivity are obtained based on the optical detection benchmark parameters. The dark field reference voltage matrix refers to the background voltage distribution generated by the dark current and readout noise of the CCD camera under completely shading conditions, including the dark field reference voltage values ​​corresponding to multiple pixel points. The reference voltage matrix refers to the measurement standard reflectivity. The voltage distribution of the plate under the same optical path, the calibrated reflectivity refers to the absolute reflectivity standard value of the standard reflective plate at a wavelength of 450nm, and then the difference between the original voltage value of each pixel and the dark field reference voltage value is calculated to obtain the calibration voltage matrix, where the calibration voltage matrix refers to the effective signal voltage distribution after deducting the dark field reference voltage value from the original voltage value of each pixel. Then, the ratio of the calibration voltage value to the reference voltage value of each pixel is calculated, and the product of the ratio and the calibrated reflectivity is calculated to obtain the reflectivity measurement value corresponding to each point, where the reflectivity measurement value refers to the absolute reflectivity percentage of each point on the fabric surface relative to the standard reflective plate. The two-dimensional array formed by arranging the reflectivity measurement values ​​of all three-dimensional points according to the spatial coordinates is the reflectivity matrix. Then the reflectivity measurement value of each pixel point and the reflectivity fluctuation prediction value predicted by the model are synchronously input into the locked amplifier, and the time constant is set according to the main frequency of vibration. Then, the reflectivity measurement value and the reflectivity fluctuation prediction value are multiplied to obtain the vibration synchronization component, where the vibration synchronization component refers to the reflectivity fluctuation signal that maintains strict phase synchronization with the vibration modulation model. This method of extracting the component in the measurement signal that is in phase with the model predicted fluctuation through orthogonal demodulation can accurately separate the reflectivity modulation component caused by vibration and improve the signal-to-noise ratio of the reflectivity measurement value. Improve by ≥15dB while retaining the spatial resolution of the true optical characteristics of the defects at no less than 10μm. Then, the vibration synchronization component is deducted from the reflectivity measurement value to obtain the anti-disturbance reflectivity distribution matrix. The anti-disturbance reflectivity distribution matrix refers to a two-dimensional steady-state optical parameter array that characterizes the true material optical properties after eliminating vibration interference. This method achieves precise separation and subtraction of the vibration interference component, effectively eliminating the reflectivity measurement error caused by high-frequency vibration of 50-500Hz. It can solve the problem of existing technologies that high-frequency mechanical vibration causes deviation in fabric reflectivity measurement values, thereby affecting the accurate detection of optical characteristics such as fabric defects; Then, the reference light intensity value and the characteristic absorption curve of the fabric are obtained according to the optical detection benchmark parameters, where the reference light intensity value refers to the stable light power density reference value calibrated using a NIST traceable standard light source under a standard detection environment, and the characteristic absorption curve of the fabric refers to the quantitative value used to characterize the light absorption characteristics of the fabric at different wavelengths. The dynamic light disturbance amount collected in real time and the reference light intensity value are input into the high-speed analog divider for hardware-level ratio calculation to generate a dynamic compensation coefficient, where the dynamic compensation coefficient refers to the light intensity correction factor calculated in real time, which is used to offset the influence of ambient light fluctuations on the measurement. This coefficient is passed The 16-bit DAC is used to feed back the signal to the preamplifier of the detection optical path in real time (gain adjustment range ±20dB), and the output signal of the photomultiplier tube is closed-loop gain controlled to ensure that under the condition of ±15% illumination fluctuation, the absolute light intensity value deviation measured by the system is ≤0.5%. Then, the spectrum offset and the reference wavelength of 450nm are input into the floating-point arithmetic unit built into the FPGA to calculate the wavelength compensation coefficient. Then, dynamic light intensity compensation is performed on the anti-interference reflectivity distribution matrix: the reflectivity value of each pixel is multiplied by the dynamic compensation coefficient of the corresponding pixel position to achieve illumination compensation. Intensity normalization correction is performed to obtain a compensated reflectance matrix, ensuring accurate gain compensation for each pixel. This point-by-point calculation based on spatial registration maintains the spatial resolution of the original image while effectively eliminating reflectance measurement deviations caused by non-uniform illumination. The matrix is ​​then spectrally corrected using wavelength compensation coefficients. The reflectance data is matched to a standard wavelength using a polynomial fitting method to generate a spectrally normalized reflectance matrix. Simultaneously, the system precisely controls the image sensor's exposure timing using a flicker synchronization trigger signal, maintaining it at an integer multiple of the light source flicker period to eliminate flicker interference. Finally, the time-synchronized reflectance matrix is ​​spatially weighted fused with the wavelength-compensated absorbance data. A high-precision analog-to-digital converter outputs standardized optical signature enhancement data with a signal-to-noise ratio improvement of over 20 dB and a spatial resolution of 10 μm. Traditional methods struggle to cope with dynamic changes in ambient illumination and spectral shifts, leading to systematic deviations in reflectance measurements. However, this solution, through a dynamic calibration system combining hardware-level real-time compensation with multi-dimensional optical parameter calibration, achieves real-time closed-loop gain control and precise correction of spectral shifts for ambient illumination fluctuations.

[0027] In one embodiment, the step S4 of acquiring fabric defect characterization data based on the optical feature enhancement data includes: S41, obtaining the RGB value of each pixel according to the optical feature enhancement data; S42, obtaining an optical feature-noise correlation database and a fabric reference spectrum; S43, obtaining the minimum detection size of the defect, and obtaining an optical characteristic component according to the RGB value corresponding to each pixel, the fabric reference spectrum, and the minimum detection size of the defect; S44, acquiring an optimized characteristic component according to the optical characteristic-noise association database and the optical characteristic component, wherein the optimized characteristic component includes a base reflection component, an abnormal absorption component, and a high-frequency scattered noise component; S45. Obtain a high-frequency texture feature map based on the fabric reference spectrum and the abnormal absorption component, and obtain a comprehensive defect characterization map based on the base reflection component, the abnormal absorption component, and the high-frequency scattered noise component.

[0028] As described in steps S41-S45 above, the present invention obtains the RGB value of each pixel through optical feature enhancement data, where the RGB value refers to the numerical value of the color and brightness of a single pixel in an image using a combination of the three primary colors of red, green, and blue. Furthermore, an optical feature-noise correlation database and a fabric reference spectrum are obtained. The optical feature-noise correlation database refers to a pre-constructed structured data set that stores the correspondence between the optical features of specific fabric defects and various types of interference noise (such as mechanical vibration, light fluctuation, etc.). The optical feature-noise correlation database can characterize the characteristic frequency band distribution of specific defects at the target wavelength, the attenuation weight coefficients of different vibration frequencies on the optical signal, and the change in contrast attenuation caused by light fluctuation with frequency. The fabric reference spectrum refers to a storage of optical characteristic parameters of fabrics of different materials, such as the normal reflectivity distribution range corresponding to yarn density. Then, combined with textile process specifications, the minimum detection size of the defect (such as a hole diameter ≥ 0.5 mm, a yarn break length ≥ 1 mm, etc.) is obtained. The RGB value of each pixel is then matched channel by channel with the fabric reference spectrum, and the initial base reflection component is obtained through least squares fitting (indicating the degree of match between the pixel and the standard fabric reflectance). Residual analysis is then used to extract the initial abnormal absorption component (i.e., the area where the pixel reflectance has a significant negative deviation from the standard value. Its spatial distribution must meet the process size threshold to be determined as a valid defect signal). High-pass filtering is also used to separate the initial high-frequency scattering noise component (whose spatial frequency characteristics do not match the texture period defined by the process parameters). These three components constitute the optical characteristic components, and the decomposition parameters must be dynamically adjusted during the separation process to ensure that the initial base reflection component at the macro scale (e.g., ≥5mm) fully preserves the overall optical properties of the fabric, while the initial abnormal absorption component at the micro scale (e.g., 0.5-1mm) accurately captures tiny defects. The final output optical characteristic components will serve as the quantitative basis for subsequent defect identification, achieving a deep coupling of "feature extraction and process specification." Then, the decomposed optical characteristic components are optimized in a targeted manner based on the optical characteristic-noise correlation database: for the initial component of the base reflection, a vibration frequency domain compensation algorithm is used. Specifically, after converting the spatial domain signal to the frequency domain through Fourier transform, the interference weight coefficient of the dominant frequency band of mechanical vibration in the mapping function of the optical characteristic-noise correlation data is used, and an inverse filtering process is performed in the frequency domain to eliminate the low-frequency baseline drift caused by vibration, thereby obtaining the base reflection component; for the initial component of the abnormal absorption, it is enhanced by a spectral matching filter. The kernel function of the filter is designed according to the standard absorption spectrum characteristics of the defect in the mapping function, and a convolution operation is performed in the spatial domain to improve the signal-to-noise ratio, while suppressing the interference components that match the frequency domain attenuation model of the light fluctuation, thereby obtaining the abnormal absorption component; for the high-frequency scattering For the initial component of the emitted noise, an optical coherence mask is constructed using the noise base characteristics in the mapping function. The incoherent noise is separated and filtered out through the wavelet threshold denoising method to obtain the high-frequency scattered noise component. The base reflection component, the abnormal absorption component and the high-frequency scattered noise component constitute the optimized feature components. The optimized feature components must meet the following requirements: the energy fluctuation of the base reflection component is ≤5%, the peak signal-to-noise ratio of the abnormal absorption component is ≥30dB, and the energy proportion of the scattered noise component is ≤1%. These optimized components will serve as input data for subsequent feature fusion. This method of separating incoherent noise through "optical coherence mask + wavelet denoising" achieves precise suppression of "one type of interference corresponds to one optimization strategy", thereby solving the problem that the existing technology lacks targeted processing of noise components. Then, based on the standard spectral database (including standard reflection / absorption spectral curves of various defects at the target wavelength) and the fabric benchmark spectrum, the optimized characteristic components are matched with the standard library data in multiple dimensions: the pixel-by-pixel Euclidean distance is calculated based on the abnormal absorption component (obtained through spectral-spatial decomposition) and the benchmark reflectivity curve in the standard spectral database, and an initial difference map is generated. Then, the guided filtering algorithm is used for edge enhancement: the yarn strike field (extracted through a Gabor filter group, with a frequency set to 10-15 lines / mm to match the fabric texture) is used as the guide image, and the edge of the difference map is smoothed. The filter radius is set to 3 pixels (corresponding to a process accuracy of 0.1mm), and the regularization parameter is set to 0.1mm. The method retains gradient changes of ≥0.5dB, and then uses nonlinear stretching (Sigmoid function, slope coefficient k=5) to enhance the reflection gradient difference at the edge of the defect, so that the sudden gradient at the yarn break (usually ≥8dB / mm) appears as a significant bright spot in the feature map, while the gentle gradient of the normal texture (≤2dB / mm) is suppressed. Finally, a high-frequency texture feature map is output, and abnormal areas with Euclidean distance exceeding the threshold and gradient amplitude greater than 6dB are marked as input data for subsequent defect location; and based on the abnormal absorption component, combined with the stain absorption parameters pre-stored in the standard spectral database, its significance score is calculated pixel by pixel, and then an adaptive threshold segmentation algorithm is used to segment the areas with a significance score exceeding 1.0 The pixel area is marked as a high-attention area, and small holes are filled through morphological closing operations (3×3 circular structural elements) to ensure the integrity of continuous defect areas. Then, bilateral filtering is used to smooth the edges of the initial weights to avoid weight jumps caused by noise. Finally, the attention weight matrix is ​​obtained. At the same time, through the cross-validation mechanism, it is required that high-weight areas must appear simultaneously in the feature pyramid of at least two adjacent scales (such as L1 and L3 layers) to eliminate isolated noise interference. Finally, multi-scale fusion and verification of the 5-level feature pyramid are realized, and a hierarchical processing architecture is constructed based on the optimized feature components: the first level uses the high-frequency gradient features in the abnormal absorption component (the signal-to-noise ratio is improved by 30dB after vibration frequency domain compensation) The method uses local binary patterns (LBP) to analyze sudden changes in the surface reflectance of a single yarn. A valid warp break signal is identified when the reflectance gradient of three consecutive pixels exceeds 8 dB and the coefficient of variation of the base reflection component in this region is ≤5%. In the second stage, a multi-directional Gabor filter bank (0° / 45° / 90°, 10 lines / mm) is applied to extract the yarn interlacing point features, combining the texture stability of the base reflection component and the spatial distribution of the abnormal absorption component. The energy proportion of the scattered noise component in this frequency band is verified to be less than 10% to eliminate interference. In the third stage, a morphological opening operation is performed to process the connected domain of the abnormal absorption component (structuring element 3×3). The SSIM value of the base reflection component in the area surrounding the defect is required to be ≥0.85 to ensure the accuracy of defect isolation; the fourth level performs a Fourier transform on the substrate reflection component to detect the peak amplitude spectrum of periodic defects (must exceed the energy of the scattered noise component by 3σ), and combines it with the spatial correlation analysis of the abnormal absorption component to confirm the authenticity of the defect; the fifth level comprehensively evaluates the overall uniformity of the substrate reflection component (SSIM ≥ 0.9) and the aggregation characteristics of the abnormal absorption component, and finally outputs a comprehensive defect characterization map that integrates multi-scale optical properties. This acquisition method can effectively eliminate isolated noise interference and ensure the accuracy of defect identification. Each defect is annotated with its source component and the cross-scale consistency verification results.

[0029] In one embodiment, the step S5 of obtaining a quality quantification index based on the high-frequency texture feature map and the defect comprehensive characterization map, and obtaining a comprehensive risk value based on the quality quantification index includes: S51, obtaining an optical signal difference value of each pixel point according to the high-frequency texture feature map, and obtaining a defect heat map according to the optical signal difference value, the attention weight matrix and the defect comprehensive representation map; S52, obtaining defect optical characterization data according to the defect thermal map, and obtaining a defect detection result table according to the defect optical characterization data; S53, using the defect heat map and the defect detection result table as quality quantification indicators; S54. Obtain a defect density distribution matrix according to the defect heat map, and obtain defect size, defect type, and defect coordinates according to the defect density distribution matrix and the defect comprehensive characterization map; S55. Obtain a defect impact index according to the defect size, defect type, and defect coordinates, and obtain a comprehensive risk value according to the defect density distribution matrix, defect type, and defect impact index.

[0030] As described in the above steps S51-S55, the present invention obtains the optical signal difference value of each pixel through the high-frequency texture feature map, wherein the optical signal difference value refers to the numerical difference between the optical signal (such as reflectivity, absorptivity, etc.) of a certain pixel point in the high-frequency texture feature map and the standard optical signal of the corresponding pixel point of the fabric reference spectrum, which is used to quantify the degree to which the optical characteristics of the pixel point deviate from the normal state, and then perform a dot multiplication operation with the spatial domain attention weight matrix, and strengthen the optical abnormal signal of the real defect area through weight adjustment (such as the absorptivity difference of the stain area is amplified by a weight of 0.9) while suppressing environmental interference (such as the difference value of the reflective area is attenuated by a weight of 0.2), and then adopts a dynamic threshold algorithm, in which the threshold is adaptively adjusted according to the weight mean of the current area (high weight area The threshold is lowered by 5-10% to capture weak signal defects, and the threshold in low-weight areas is increased by 20% to filter texture noise. Finally, a morphological opening operation is performed on the weighted difference map to eliminate discrete noise points, and hot spots with a continuous area of ​​more than 3 pixels are marked. After normalization, multi-level consistency is verified using a 5-scale feature pyramid (for example, a broken warp hot spot detected in the L1 layer must have an extension feature in the L3 layer; otherwise, it is judged as an artifact). Finally, a defect heat map that strictly corresponds to the physical defects of the fabric is output. This method strengthens the effective signal through "differential weight adjustment" and adopts a dynamic threshold method based on the weighted mean to achieve a balance between "weak signal defect capture" and "texture noise filtering", avoiding the drawbacks of traditional fixed thresholds such as "missing weak signals" or "falsely detecting noise". Based on the defect heat map, the coordinates of the abnormal area are obtained, and the optical characterization data of the defects at the corresponding positions are extracted for quantitative analysis: for each high-value area of ​​the heat map, its optical response mode is matched through a 5-scale feature pyramid (for example, at a scale of 50μm, a warp break is manifested as a sudden drop in reflectivity ≥12dB and the direction of the high-frequency gradient is consistent with the yarn direction; at a scale of 1mm, the stain presents an absorption rate peak at a specific wavelength exceeding the standard value by 15%). At the same time, the spatial attention weight matrix is ​​combined to verify the feature significance, and then the standard spectral database is called for spectral angle matching to calculate the spectral curve of the defect area and the standard defect. The system measures the similarity of the defect templates and records the core optical parameters of each defect, including transmittance, reflection intensity and spectral feature similarity. Finally, it compares the core optical parameters of each defect with the process threshold of the textile material optical standard library to generate a defect detection result table. Each record in the defect detection result table contains the defect mark, center coordinates, core optical parameters, scale level association identifier, etc. The defect detection result table can be directly imported into the quality management system for defect tracing and process optimization, helping to quickly troubleshoot equipment or algorithm parameter problems, thereby solving the problem that existing technologies cannot locate the source of detection deviation during defect tracing. The defect heat map is then gridded, and the fabric is divided into 5mm×5mm detection units. The area ratio and number of defects in the high-value areas of the heat map in each unit are counted to generate a defect density distribution matrix, which directly reflects the optical uniformity defect distribution of the fabric. At the same time, three core parameters, namely defect size, defect type and defect coordinates, are obtained from the defect detection result table: First, based on the multi-level detection results of the 5-scale feature pyramid, cross-scale pixel aggregation is performed on each confirmed defect area: starting from the highest resolution L1 layer, optical abnormal areas (such as continuous pixel blocks with a sudden drop in reflectivity exceeding the threshold) are marked through connected domain analysis, and abnormal areas with corresponding spatial positions are searched in adjacent scale layers (L2-L5) (a pyramid coordinate mapping algorithm is used to ensure cross-scale position alignment). Then, the connected domain pixels detected at each level are accumulated according to the weights through a multi-scale fusion algorithm, and duplicate counting caused by differences in detection sensitivity at different scales is eliminated (such as the broken warp image marked in the L1 layer). Pixels are no longer recalculated in the L3 layer). Physical size conversion is then performed based on preset camera resolution parameters. For irregularly shaped defects, the minimum enclosing rectangle algorithm is used to determine the characteristic size (major axis × minor axis), ultimately outputting the defect size for each defect. Based on the defect optical characterization data, the spectral response value, reflectance and absorption intensity change values, and frequency domain energy distribution characteristics (wavelet analysis is used to obtain the high-frequency / mid-frequency energy ratio) corresponding to each defect area are extracted. A pre-trained random forest classifier is then applied to determine the defect type, achieving precise mapping from optical features to defect type and avoiding misclassifications caused by traditional single-feature classification (such as grayscale analysis). Finally, the heat map coordinates are mapped to the physical location of the fabric to obtain the defect coordinates. A weighted summation algorithm is then used to calculate the defect impact index based on the defect size, defect type, and defect coordinates. The defect impact index is a quantitative indicator used to assess the impact of a defect on the overall fabric quality, laying the foundation for subsequent comprehensive risk assessment. Finally, the defect density distribution matrix, defect type, and defect impact index are normalized: the defect density distribution is converted to a relative density value in the range of 0-1, the defect impact index is mapped to a numerical value according to the defect level, and the defect type weight is directly used as the coefficient. The comprehensive risk value is then calculated using a weighted fusion algorithm. At the same time, a spatial distribution correction factor is introduced, and an additional 20% risk weight is added to continuous defect areas (area >10mm²). The final output comprehensive risk value is compressed to the range of 0-1 using a Sigmoid function. This multi-dimensional risk assessment method can improve the accuracy of defect risk assessment and ensure the comparability of risk values ​​across different batches and fabrics, thus solving the problem of existing technologies that have no unified risk value scale and cannot compare quality levels horizontally.

[0031] In one embodiment, the step S6 of generating hierarchical response control information according to the quality quantification indicator includes: S61. Obtaining defect characteristic energy and defect diffusion rate according to the quality quantification index, and obtaining high-frequency noise energy according to the environmental interference data; S62, obtaining optical characteristic stability according to the high-frequency noise energy and the defect characteristic energy, and determining whether the optical characteristic stability is greater than a stability threshold; S63, if the optical characteristic stability is greater than the stability threshold, obtaining the optical characteristic parameters of the defect according to the defect comprehensive characterization map; S64, obtaining a fabric optical property reference library, and obtaining a reflectivity anomaly amplitude, a transmittance gradient change, and an absorption peak offset based on the fabric optical property reference library and the optical characteristic parameters; S65, obtaining an optical sensitivity level according to the reflectivity abnormality amplitude, the transmittance gradient change, and the absorption peak offset; S66: Acquire a graded response database, and acquire graded response control information according to the graded response database, the optical sensitivity level, and the defect diffusion rate.

[0032] As described in steps S61-S66 above, the present invention obtains defect feature energy by performing energy integration on the pixel values ​​of the defect area (located by the defect heat map) in the high-frequency texture feature map, wherein the defect feature energy refers to a value reflecting the intensity of the optical feature mutation caused by the defect, and extracts the invalid optical signal energy caused by mechanical vibration and environmental interference from the environmental interference data. After converting the time domain vibration data into a frequency domain energy distribution through fast Fourier transform, the square of the amplitude of a preset high-frequency band (such as 50-1000 Hz) is accumulated to obtain high-frequency noise energy, wherein the high-frequency noise energy refers to a value used to quantify the degree of contamination of the optical detection system by external interference. At the same time, the ratio of the high-frequency noise energy to the defect feature energy is then calculated, and the difference between "1" and the ratio is calculated to obtain the optical feature stability, wherein the optical feature stability refers to an indicator reflecting the reliability of the defect detection result under environmental interference. The higher the value, the lower the degree of contamination of the effective optical feature by the high-frequency noise, and the more reliable the detection result. This method of converting the "impact of environmental interference on detection results" into a quantifiable indicator can avoid the ambiguity of the traditional "subjective judgment of interference intensity". The optical feature stability is then compared with the stability threshold. If the optical feature stability is less than the stability threshold, the current optical signal is determined to be unreliable and a re-inspection is required. That is, the imaging module is controlled to quickly switch to three characteristic wavelengths (for example, 450nm blue light, 550nm green light, and 850nm near-infrared), covering the main optical response bands of the fabric, to obtain new optical response data. Multi-spectral complementarity can identify defects that are difficult to detect with a single wavelength, thereby solving the problem of insufficient optical contrast of certain defects at specific wavelengths. The newly acquired optical response data replaces the original optical response data to execute subsequent steps, realizing a closed-loop control of "unreliable signal-automatic re-inspection-data replacement". If the optical feature stability is greater than the stability threshold, the current optical signal is determined to be reliable. This method of accurately identifying signal reliability through stability indicators is conducive to reducing the false positive rate. Then, the optical characteristic parameters of the defects are extracted from the defect detection result table. The optical characteristic parameters of the defects refer to the key physical quantities used to quantify the difference in optical characteristics between the defects and the normal fabric areas, including the reflectivity of the defect area, the transmittance of the defect area and the absorption peak wavelength of the defect area. The reflectivity of the defect area refers to the percentage ratio of the surface reflected light intensity to the incident light intensity of the defect area under the conditions of a specific wavelength of light, which is used to quantify the abnormal reflection characteristics caused by the defect (such as the reflectivity of the broken warp defect is significantly reduced due to the yarn breakage). The transmittance of the defect area refers to the ability of the defect area to transmit light of a specific wavelength, which is used to reflect the change in light transmittance caused by the defect (such as the transmittance of the hole defect is significantly reduced due to the lack of structure). Increase), the absorption peak wavelength of the defect area refers to the characteristic wavelength position where the defect area absorbs the strongest light energy within a specific spectral band, which is used to characterize the characteristic absorption changes caused by pollutants or structural defects. Then, the fabric optical property benchmark library is obtained, where the fabric optical property benchmark library refers to the optical parameter data set of defect-free fabric samples collected under standard laboratory conditions in multi-spectral bands (such as 450nm / 550nm / 850nm), including reflectance benchmark value, transmittance benchmark value and absorption peak benchmark wavelength, and the defect area reflectance, defect area transmittance and defect area absorption peak wavelength are compared with the reflectance benchmark value, transmittance benchmark value and absorption peak benchmark wavelength respectively. By comparison, we can obtain the reflectivity anomaly amplitude (the absolute difference percentage between the reflectivity of the defect area and the reflectivity reference value at the same wavelength, which is used to quantify the degree of attenuation or enhancement of reflectivity caused by the defect), the transmittance gradient change (the maximum change rate of the transmittance of the defect area in the spatial dimension, which reflects the sudden change intensity of the transmittance characteristics of the defect boundary) and the absorption peak offset (the absolute difference between the absorption peak wavelength of the defect area and the absorption peak reference wavelength, which directly characterizes the spectral characteristic displacement caused by pollutants or chemical degeneration). Through these three core indicators, we can achieve full-dimensional quantification of the optical property differences of the defects, avoiding the one-sidedness of the traditional single parameter (such as only looking at the reflectivity). Then, through the first-level nodes of the decision tree, we can determine the difference in the optical properties of the defects. Defects are roughly classified based on their dominant optical characteristics: if the absorption peak offset is greater than 15nm, they are classified as oil stains; if the reflectivity anomaly is greater than 25%, they are classified as broken warps; otherwise, if the transmittance gradient change is greater than 4%, they are classified as holes. A dedicated fuzzy logic rule base is then applied to each defect type at the sub-level node to determine the optical sensitivity level. The optical sensitivity level is a risk quantification indicator based on the severity of the defect's optical characteristics and its impact on product quality. This precise mapping method from parameter quantification to level classification ensures that the optical sensitivity level matches the defect's actual quality impact. Then, based on the coordinates of the abnormal area, the optical flow method is used to calculate the boundary displacement of continuous time frames (such as 10 frames per second) to obtain the pixel displacement of the defect area in the longitude and latitude directions (such as +5 pixels / frame in the X direction and +2 pixels / frame in the Y direction). Combined with the real-time linear speed, the pixel displacement is converted into a physical diffusion rate (such as 0.5mm / s in the X direction and 0.2mm / s in the Y direction). Finally, the defect diffusion rate is obtained through vector synthesis. The defect diffusion rate refers to the physical expansion speed of the defect area on the fabric surface per unit time (such as per second). This method of extending the "static characteristics of defects" to the "dynamic diffusion trend" can provide a "timely" basis for production line control, avoiding the lag of traditional methods that only formulate control strategies based on static defects. And further obtain a graded response database, where the graded response database refers to a rule base constructed based on textile production line process knowledge and used to associate optical detection results with production line control strategies. The rule base stores the mapping relationship between optical sensitivity level, defect diffusion rate and control parameters in a tabular form, and then inputs the optical sensitivity level and defect diffusion rate in real time. Based on the graded response database, the optimal control strategy is retrieved through interval matching to obtain graded response control information. The graded response control information contains JSON format control instructions for the speed reduction ratio and light source adjustment parameters, thereby converting abstract risk indicators into executable production line adjustment operations to achieve seamless connection of "detection-decision-control", which is conducive to meeting the real-time and standardization requirements of the production line.

[0033] This application also provides an intelligent fabric defect detection system based on AI visual recognition, including: The data acquisition module is used to obtain optical response data of the fabric surface using visible light of a specific wavelength based on AI visual recognition, and obtain the fabric's motion state data and environmental interference data through sensors; a motion correction module, configured to obtain primary correction data based on the motion state data and the optical response data; a parameter calibration module, configured to perform dynamic optical parameter calibration on the primary correction data according to the environmental interference data to obtain optical feature enhancement data; a defect analysis module, configured to obtain fabric defect characterization data based on the optical feature enhancement data, wherein the fabric defect characterization data includes a high-frequency texture feature map and a defect comprehensive characterization map; a detection and evaluation module, configured to obtain a quality quantitative index based on the high-frequency texture feature map and the defect comprehensive characterization map, and obtain a comprehensive risk value based on the quality quantitative index; A detection and control module is used to determine whether the comprehensive risk value is greater than a preset threshold; If the comprehensive risk value is not greater than a preset threshold, outputting the quality quantitative index; If the comprehensive risk value is greater than a preset threshold, graded response control information is generated according to the quality quantification index, and adjustment operations are performed on the textile production line according to the graded response control information, wherein the graded response control information includes a speed reduction ratio and a light source adjustment parameter.

[0034] In one embodiment, the motion correction module includes: A motion sensing unit, configured to obtain a motion speed parameter, a motion direction, and a motion period of the fabric according to the motion state data, and to obtain a motion blur length of the fabric according to the motion speed parameter; a fuzzy quantization unit, configured to obtain an optical detection distance and a static fuzzy base value, and obtain an optical fuzzy characteristic coefficient according to the optical detection distance, the static fuzzy base value, and the motion blur length; A matrix construction unit, configured to obtain a time-varying blur kernel matrix according to the optical blur characteristic coefficient and the fabric movement direction; a subframe interception unit, configured to obtain an overlapping subframe sequence according to the optical response data and the fabric motion period; a noise evaluation unit, configured to obtain a vibration sensitivity coefficient and obtain a dynamic noise parameter according to the vibration sensitivity coefficient and the motion speed parameter; a noise suppression unit, configured to obtain a defect characteristic response spectrum, and obtain a denoised subframe sequence according to the defect characteristic response spectrum, the overlapping subframe sequence and the dynamic noise parameter; A data correction unit is configured to obtain a deblurred subframe sequence according to the motion speed parameter, the time-varying blur kernel matrix and the denoised subframe sequence, and to obtain primary correction data according to the fabric motion period and the deblurred subframe sequence.

[0035] The present invention also provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned intelligent detection method for fabric defects based on AI visual recognition are implemented.

[0036] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of an intelligent method for detecting fabric defects based on AI visual recognition.

[0037] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided in this application and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct RAM bus dynamic RAM (DRDRAM), and RAM bus dynamic RAM (RDRAM).

[0038] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, apparatus, article, or method comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, apparatus, article, or method. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, apparatus, article, or method comprising the element.

[0039] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. The intelligent fabric defect detection method based on AI visual recognition is characterized by: include: Based on AI visual recognition, it uses visible light of a specific wavelength to obtain optical response data on the fabric surface, and uses sensors to obtain fabric motion state data and environmental interference data; acquiring primary correction data according to the motion state data and the optical response data; Performing dynamic optical parameter calibration on the primary correction data according to the environmental interference data to obtain optical feature enhancement data; Acquiring fabric defect characterization data based on the optical feature enhancement data, wherein the fabric defect characterization data includes a high-frequency texture feature map and a defect comprehensive characterization map; Obtaining a quality quantification index based on the high-frequency texture feature map and the defect comprehensive characterization map, and obtaining a comprehensive risk value based on the quality quantification index; Compare the comprehensive risk value with the preset threshold: If the comprehensive risk value is not greater than a preset threshold, outputting the quality quantitative index; If the comprehensive risk value is greater than a preset threshold, graded response control information is generated according to the quality quantification index, and adjustment operations are performed on the textile production line according to the graded response control information, wherein the graded response control information includes a speed reduction ratio and a light source adjustment parameter.

2. The fabric defect intelligent detection method based on AI visual recognition according to claim 1 is characterized in that: The step of acquiring primary correction data according to the motion state data and the optical response data comprises: Acquire a motion speed parameter, a motion direction, and a motion period of the fabric according to the motion state data, and acquire a motion blur length of the fabric according to the motion speed parameter; Obtaining an optical detection distance and a static blur base value, and obtaining an optical blur characteristic coefficient according to the optical detection distance, the static blur base value and the motion blur length; Obtaining a time-varying blur kernel matrix according to the optical blur characteristic coefficient and the fabric movement direction; Acquire a sequence of overlapping subframes according to the optical response data and the fabric motion period; Obtaining a vibration sensitivity coefficient, and obtaining a dynamic noise parameter according to the vibration sensitivity coefficient and the motion speed parameter; Acquire a defect characteristic response spectrum, and acquire a denoised subframe sequence according to the defect characteristic response spectrum, the overlapping subframe sequence and the dynamic noise parameter; A deblurred subframe sequence is obtained according to the motion speed parameter, the time-varying blur kernel matrix and the denoising subframe sequence, and primary correction data is obtained according to the fabric motion period and the deblurred subframe sequence.

3. The fabric defect intelligent detection method based on AI visual recognition according to claim 1 is characterized in that: The step of performing dynamic optical parameter calibration on the primary correction data according to the environmental interference data to obtain optical feature enhancement data includes: Acquiring mechanical vibration parameters and ambient light parameters according to the environmental interference data; Obtaining a calibration conversion coefficient, and obtaining a reflectivity fluctuation prediction value based on the calibration conversion coefficient and the mechanical vibration parameter; Acquiring optical detection reference parameters, and performing phase-sensitive demodulation processing on the primary correction data according to the optical detection reference parameters and the reflectivity fluctuation prediction value to obtain an anti-disturbance reflectivity distribution matrix; Obtaining a dynamic light disturbance amount and a spectral shift amount according to the ambient light parameter, and obtaining a dynamic compensation coefficient and a wavelength compensation coefficient according to the dynamic light disturbance amount, the spectral shift amount and the optical detection reference parameter; Optical feature fusion is performed on the primary correction data according to the anti-disturbance reflectivity distribution matrix, the dynamic compensation coefficient and the wavelength compensation coefficient to obtain optical feature enhancement data.

4. The fabric defect intelligent detection method based on AI visual recognition according to claim 1 is characterized in that: The step of acquiring fabric defect characterization data based on the optical feature enhancement data includes: Obtaining the RGB value of each pixel according to the optical feature enhancement data; Obtaining an optical feature-noise correlation database and fabric benchmark spectra; Obtaining a minimum detection size of a defect, and obtaining an optical characteristic component based on the RGB value corresponding to each pixel, a fabric reference spectrum, and the minimum detection size of the defect; Acquire an optimized characteristic component according to the optical characteristic-noise correlation database and the optical characteristic component, wherein the optimized characteristic component includes a substrate reflection component, an abnormal absorption component, and a high-frequency scattered noise component; A high-frequency texture feature map is obtained based on the fabric reference spectrum and the abnormal absorption component, and a comprehensive defect characterization map is obtained based on the base reflection component, the abnormal absorption component and the high-frequency scattering noise component.

5. The fabric defect intelligent detection method based on AI visual recognition according to claim 1 is characterized in that: The step of obtaining a quality quantification index based on the high-frequency texture feature map and the defect comprehensive characterization map, and obtaining a comprehensive risk value based on the quality quantification index includes: Obtaining an optical signal difference value for each pixel according to the high-frequency texture feature map, and obtaining a defect heat map according to the optical signal difference value, the attention weight matrix, and the defect comprehensive representation map; Obtaining defect optical characterization data according to the defect thermal map, and obtaining a defect detection result table according to the defect optical characterization data; Using the defect heat map and the defect detection result table as quality quantitative indicators; Obtaining a defect density distribution matrix according to the defect heat map, and obtaining defect size, defect type, and defect coordinates according to the defect density distribution matrix and the defect comprehensive characterization map; A defect impact index is obtained according to the defect size, defect type and defect coordinates, and a comprehensive risk value is obtained according to the defect density distribution matrix, defect type and the defect impact index.

6. The fabric defect intelligent detection method based on AI visual recognition according to claim 1 is characterized in that: The step of generating hierarchical response control information according to the quality quantification indicator includes: Obtaining defect characteristic energy and defect diffusion rate according to the quality quantification index, and obtaining high-frequency noise energy according to the environmental interference data; Obtaining optical characteristic stability according to the high-frequency noise energy and the defect characteristic energy, and determining whether the optical characteristic stability is greater than a stability threshold; If the optical characteristic stability is greater than the stability threshold, obtaining the optical characteristic parameters of the defect according to the defect comprehensive characterization map; Obtaining a fabric optical property reference library, and obtaining a reflectivity anomaly amplitude, a transmittance gradient change, and an absorption peak offset based on the fabric optical property reference library and the optical characteristic parameters; Obtaining an optical sensitivity level according to the reflectivity anomaly amplitude, the transmittance gradient change, and the absorption peak shift; A graded response database is obtained, and graded response control information is obtained according to the graded response database, the optical sensitivity level, and the defect diffusion rate.

7. The fabric defect intelligent detection system based on AI visual recognition is characterized by: include: The data acquisition module is used to obtain optical response data of the fabric surface using visible light of a specific wavelength based on AI visual recognition, and obtain the fabric's motion state data and environmental interference data through sensors; a motion correction module, configured to obtain primary correction data based on the motion state data and the optical response data; a parameter calibration module, configured to perform dynamic optical parameter calibration on the primary correction data according to the environmental interference data to obtain optical feature enhancement data; a defect analysis module, configured to obtain fabric defect characterization data based on the optical feature enhancement data, wherein the fabric defect characterization data includes a high-frequency texture feature map and a defect comprehensive characterization map; a detection and evaluation module, configured to obtain a quality quantitative index based on the high-frequency texture feature map and the defect comprehensive characterization map, and obtain a comprehensive risk value based on the quality quantitative index; A detection and control module is used to determine whether the comprehensive risk value is greater than a preset threshold; If the comprehensive risk value is not greater than a preset threshold, outputting the quality quantitative index; If the comprehensive risk value is greater than a preset threshold, graded response control information is generated according to the quality quantification index, and adjustment operations are performed on the textile production line according to the graded response control information, wherein the graded response control information includes a speed reduction ratio and a light source adjustment parameter.

8. The fabric defect intelligent detection system based on AI visual recognition according to claim 7 is characterized in that: The motion correction module includes: A motion sensing unit, configured to obtain a motion speed parameter, a motion direction, and a motion period of the fabric according to the motion state data, and to obtain a motion blur length of the fabric according to the motion speed parameter; a fuzzy quantization unit, configured to obtain an optical detection distance and a static fuzzy base value, and obtain an optical fuzzy characteristic coefficient according to the optical detection distance, the static fuzzy base value, and the motion blur length; A matrix construction unit, configured to obtain a time-varying blur kernel matrix according to the optical blur characteristic coefficient and the fabric movement direction; a subframe interception unit, configured to obtain an overlapping subframe sequence according to the optical response data and the fabric motion period; a noise evaluation unit, configured to obtain a vibration sensitivity coefficient and obtain a dynamic noise parameter according to the vibration sensitivity coefficient and the motion speed parameter; a noise suppression unit, configured to obtain a defect characteristic response spectrum, and obtain a denoised subframe sequence according to the defect characteristic response spectrum, the overlapping subframe sequence and the dynamic noise parameter; A data correction unit is configured to obtain a deblurred subframe sequence according to the motion speed parameter, the time-varying blur kernel matrix and the denoised subframe sequence, and to obtain primary correction data according to the fabric motion period and the deblurred subframe sequence.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images