An intelligent detection method for cloth defects based on image recognition analysis

Abstract

The application discloses a kind of cloth defect intelligent detection methods based on image recognition analysis, it is related to cloth defect detection technical field, solve the technical problem in prior art, cloth processing cannot be analyzed to accurately collect cloth reference image in historical processing process, specifically, reference image detection, image acquisition is carried out to historical completed processing cloth and whether current historical processing cloth is suitable as defect detection reference cloth is detected according to the inference of acquisition image;Defect cause evaluation, according to real-time cloth processing environment and processing technology are analyzed, the cause of the generation of cloth defect in current cloth processing process is inferred;Reference image selection, according to the selected defect detection reference cloth is processed process simulation, and according to simulated process is generated image acquisition;Defect real-time comparison, real-time production cloth is compared with defect detection reference cloth and picture analysis processing, whether current production cloth exists defect is inferred by picture comparison.

Description

Technical Field

[0001] This invention relates to the field of fabric defect detection technology, specifically to an intelligent fabric defect detection method based on image recognition analysis. Background Technology

[0002] Fabric defect detection is a crucial step in the textile industry, ensuring product quality. Various defects may occur during fabric production, such as broken warp and weft threads, thick knots, oil stains, and holes. If these defects are not detected, they will affect the quality of the final product. Different customers have different requirements for fabric quality. Detection also helps improve production efficiency and reduce costs.

[0003] Patent CN115330795A discloses a method for detecting fabric burr defects. The method includes: obtaining a judgment degree; classifying pixels into single-pixel noise and suspected noise based on the judgment degree; dividing the grayscale image into multiple different sub-regions, each with a different background grayscale value; obtaining the background grayscale feature value of each sub-region; reducing the complexity of fabric burr detection, making the burrs appear more complete and clear in the image, removing the influence of noise, bringing convenience to researchers and inspectors, and making the detection of burr defects more complete and accurate.

[0004] However, in existing technologies, fabric processing cannot analyze historical processing processes to accurately collect fabric reference images, nor can it accurately collect real-time processing defects through comparison with reference objects. Furthermore, it cannot collect real-time processed fabric sample images, resulting in deviations in real-time fabric defect detection and making it impossible to accurately detect defects. At the same time, it cannot predict defects in advance, reducing the efficiency of defect protection during processing and easily causing processing defects to occur.

[0005] To address the aforementioned technical shortcomings, a solution is proposed. Summary of the Invention

[0006] The purpose of this invention is to solve the problems mentioned above by proposing an intelligent detection method for fabric defects based on image recognition analysis.

[0007] The objective of this invention can be achieved through the following technical solutions:

[0008] A smart method for detecting fabric defects based on image recognition analysis, comprising the following steps:

[0009] Based on the reference image detection, images of historically processed fabrics are acquired, and the acquisition images are used to infer whether the current historically processed fabrics are suitable as reference fabrics for defect detection.

[0010] Defect cause assessment involves analyzing the real-time fabric processing environment and processes to infer the causes of fabric defects during the current fabric processing.

[0011] Select a reference image, simulate the processing steps based on the selected defect detection reference fabric, and generate images based on the simulated process.

[0012] Real-time defect comparison involves analyzing images of the fabric being produced in real time and comparing them with those of fabrics used for defect detection. This comparison allows for the inference of whether any defects exist in the fabric currently being produced.

[0013] As a preferred embodiment of the present invention, the image detection process is as follows:

[0014] The monitoring videos of historical fabric processing are collected, and arbitrary segments are extracted from the inspection videos based on the historical fabric processing results to obtain the processing time periods of defect-free fabric. The surface of the fabric within the corresponding processing time period is then captured, with the fabric width based on the fabric production width and the fabric length manually set by the number of images captured per unit time by the real-time production line image acquisition equipment to ensure that the captured images meet the clarity requirements.

[0015] In a preferred embodiment of the present invention, the surface of the fabric is divided into several sub-regions, and a detection image of the fabric surface is constructed by selecting a number of sub-regions. Specifically, the pixel count of each sub-region in the detection image is set to X×Y. The grayscale value of the detection image is collected by a sensor, and H(x,y) is defined as the grayscale value at (x,y) in the sub-region within the detection image. The average grayscale value of the detection image is obtained using a formula. The formula is: ;

[0016] After obtaining the average gray value, the standard deviation of the gray value of the detected image is obtained using the formula: .

[0017] In a preferred embodiment of the present invention, if the average gray value and the standard deviation of the gray value of the corresponding sub-region are both within the corresponding set threshold range, the corresponding sub-region is marked as a defect-free sub-region; if either the average gray value or the standard deviation of the gray value of the corresponding sub-region is not within the corresponding set threshold range, the corresponding sub-region is marked as a defective sub-region; the surface of the collected fabric is segmented according to the coverage area of ​​the detected image as the segmentation parameter to obtain a real-time reference fabric, and a reference fabric image is obtained based on the real-time reference fabric.

[0018] As a preferred embodiment of the present invention, the defect cause evaluation process is as follows:

[0019] Based on historical processing time data collected from the current fabric production line, and by collecting data on fabric defect types from those historical processing time periods, the defects were categorized into external force wear and internal wear. Information on external force modifications and internal wear adjustments was obtained and compared with thresholds for changes in movement trajectory spacing and excess stretch span, respectively.

[0020] If the external force modification information does not exceed the movement trajectory change spacing threshold, and the internal adjustment information does not exceed the stretching span value excess threshold, then the type other than the current type is taken as the wear type; if the external force modification information exceeds the movement trajectory change spacing threshold, or the internal adjustment information exceeds the stretching span value excess threshold, then the current type is taken as the wear type.

[0021] In a preferred embodiment of the present invention, the external force modification information and the internal production adjustment information are respectively the distance of the movement trajectory change of the moving parts corresponding to the internal and external force wear type defects in the real-time production line, and the amount by which the transverse and longitudinal stretching span of the fabric corresponding to the internal production wear type defects in the real-time production line exceeds the maximum stretching span value of the current fabric material toughness.

[0022] As a preferred embodiment of the present invention, the image selection process is as follows:

[0023] During the current production line's fluctuating production rate based on actual production speed, production line impact data and light source impact data were obtained; and these were compared with the numerical ratio threshold and the maximum reciprocating deviation threshold, respectively.

[0024] If the production line impact data exceeds the numerical ratio threshold, and the light source impact data does not exceed the maximum reciprocating deviation threshold, then the image collected in the current time period will be used as the real-time reference image sample; if the production line impact data does not exceed the numerical ratio threshold, or the light source impact data exceeds the maximum reciprocating deviation threshold, then the image collected in the current time period will not be used as the real-time reference image sample, and additional acquisition equipment will be added or the current acquisition equipment will be adjusted in subsequent acquisitions.

[0025] In a preferred embodiment of the present invention, the production line impact data and the light source impact data are respectively the ratio of the rate of decrease in the number of missed frames per unit time during the current image acquisition device's image acquisition to the sharpness deviation value of adjacent frames, and the maximum cyclic deviation value of the luminous flux received by the current light source in the fabric area of ​​the corresponding acquired image.

[0026] As a preferred embodiment of the present invention, the real-time defect comparison process is as follows:

[0027] The system compares the real-time reference image sample with the reference fabric image to obtain texture defect information and brightness defect information, and compares them with the distance ratio threshold and brightness deviation threshold respectively. If the texture defect information exceeds the distance ratio threshold and the brightness defect information does not exceed the brightness deviation threshold, it is inferred that the current comparison has no defect risk. If the texture defect information does not exceed the distance ratio threshold or the brightness defect information exceeds the brightness deviation threshold, it is inferred that the current comparison has a defect risk.

[0028] In a preferred embodiment of the present invention, the texture defect information and the brightness defect information are respectively the ratio of the overlapping distance to the non-overlapping distance of the fabric texture trend trajectory at the same position within the corresponding fabric area of ​​the reference fabric image and the sample extracted image, and the numerical deviation value of the image display brightness at any position of the reference fabric image and the sample extracted image.

[0029] Compared with the prior art, the beneficial effects of the present invention are:

[0030] 1. In this invention, images of historically processed fabrics are acquired and the current historically processed fabrics are inferred from the acquired images to determine whether they are suitable as reference fabrics for defect detection. This allows for defect detection of real-time processed fabrics using images of suitable types of exhibited fabrics, improving the efficiency of image recognition and comparison and the feasibility of fabric defect detection. This directly ensures the production line efficiency of the fabric production line, enables timely detection of fabric defects, reduces raw material losses caused by fabric defects, and improves the overall operating efficiency of the fabric production line.

[0031] By analyzing the real-time fabric processing environment and processes, the causes of fabric defects during the current processing can be inferred. The background of fabric defect formation can be accurately inferred, and targeted production control can be implemented. In addition, the influencing factors of fabric defects can be controlled during image recognition, fundamentally solving the problem of fabric processing defects, improving fabric production efficiency, and effectively controlling the defect rate of finished fabrics.

[0032] 2. In this invention, the processing steps are simulated based on the selected defect detection reference fabric, and images are generated based on the simulated process to collect surface images of each step of the fabric processing. Based on these images, it is possible to identify and infer whether there are defects in the fabric being produced, thereby improving defect detection efficiency and saving the cost of manual defect detection.

[0033] By analyzing images of fabrics produced in real time and those used for defect detection, the system can infer whether there are defects in the fabrics currently being produced. This allows for the rapid location of defects and timely repairs. It also enables the tracing of the source of defects, reducing the impact of defects on the fabric production line and preventing a decrease in production efficiency. Furthermore, it effectively controls the frequency of defects and improves the quality of fabric production. Attached Figure Description

[0034] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0035] Figure 1 This is a flowchart illustrating the overall method of the present invention;

[0036] Figure 2 This is a flowchart of the reference image detection method of the present invention. Detailed Implementation

[0037] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0039] Please see Figure 1 As shown, an intelligent defect detection method for fabric based on image recognition analysis is described. The specific defect detection method steps are as follows:

[0040] By using image detection, images of historically processed fabrics are acquired, and the acquisition images are used to infer whether the current historically processed fabrics are suitable as reference fabrics for defect detection. This allows for the use of images of suitable types of exhibited fabrics to detect defects in real-time processed fabrics, improving the efficiency of image recognition and comparison and the feasibility of fabric defect detection. This directly ensures the production line efficiency of the fabric production line, enables the timely detection of fabric defects, reduces raw material losses caused by fabric defects, and improves the overall operating efficiency of the fabric production line.

[0041] Defect cause assessment analyzes the real-time fabric processing environment and process to infer the causes of fabric defects during the current processing, accurately infers the background of fabric defect formation, and enables targeted production control. It can also control the influencing factors of fabric defects during image recognition, fundamentally solving the problem of fabric processing defects, improving fabric production efficiency, and effectively controlling the defect rate of finished fabrics.

[0042] By selecting a reference image, the processing steps of the selected defect detection reference fabric are simulated, and images are acquired based on the simulated process. This allows for the acquisition of surface images of each step in the fabric processing, enabling the identification and inference of whether defects exist in the currently produced fabric based on the images. This improves defect detection efficiency and saves the cost of manual defect detection.

[0043] Real-time defect comparison involves analyzing images of fabrics produced in real time and those with defects detected as a reference. By comparing the images, it can be inferred whether there are defects in the fabrics currently being produced. This allows for the rapid location of the defects and timely repair. It also enables the tracing of the source of defects, reducing the impact of defects on the fabric production line and preventing a decrease in production efficiency. At the same time, it can effectively control the frequency of defects and improve the quality of fabric production.

[0044] Please refer to Figure 2 As shown, the image detection process is as follows:

[0045] The monitoring video of the historical fabric processing process is collected, and the detection video is arbitrarily cut according to the historical fabric processing results to obtain the processing time period of defect-free fabric. The surface of the fabric in the corresponding processing time period is collected, and the fabric width is based on the fabric production width. The fabric length is manually set by the number of images collected per unit time by the real-time production line image acquisition equipment to ensure that the collected images can meet the clarity requirements.

[0046] The fabric surface is divided into several sub-regions, and a detection image of the fabric surface is constructed by selecting the number of sub-regions. Specifically, the pixel dimensions of each sub-region in the detection image are set to X×Y. Grayscale values ​​of the detection image are collected using a sensor, and H(x,y) is defined as the grayscale value at (x,y) within the sub-region of the detection image. The average grayscale value of the detection image is then obtained using a formula. The formula is: ;

[0047] After obtaining the average gray value, the standard deviation of the gray value of the detected image is obtained using the formula: ;

[0048] When X×Y is the set of pixels in the corresponding sub-region of the detected image, then B and BZ are the average gray value and standard deviation of the corresponding sub-region, respectively;

[0049] Compare the average gray value and standard deviation of the corresponding sub-region with the corresponding set threshold range:

[0050] If the average gray value and the standard deviation of the gray value of the corresponding sub-region are both within the corresponding set threshold range, the corresponding sub-region is marked as a defect-free sub-region; if either the average gray value or the standard deviation of the gray value of the corresponding sub-region is not within the corresponding set threshold range, the corresponding sub-region is marked as a defective sub-region.

[0051] The surface of the collected fabric is segmented using the coverage area of ​​the detected image as a selection parameter to obtain a real-time reference fabric, and a reference fabric image is obtained based on the real-time reference fabric.

[0052] The defect cause assessment process is as follows:

[0053] Based on historical processing time data collected from the current fabric production line, and by collecting data on fabric defect types from those historical processing time periods, defects are categorized into external force wear types and internal production wear types. The movement trajectory change intervals corresponding to external force wear type defects in the real-time fabric production line are obtained. The movement trajectory change interval is obtained by measuring the distance between any point on the movement trajectory after movement and the current point on the trajectory. Simultaneously, the excess of the transverse and longitudinal tensile spans of the fabric corresponding to internal production wear type defects in the real-time fabric production line compared to the current fabric material's maximum tensile span value is obtained. The movement trajectory change intervals corresponding to external force wear type defects in the real-time fabric production line, and the excess of the transverse and longitudinal tensile spans of the fabric corresponding to internal production wear type defects in the real-time fabric production line compared to the current fabric material's maximum tensile span value are marked as external force modification information and internal production adjustment information, respectively, and compared with movement trajectory change interval thresholds and tensile span value excess thresholds, respectively.

[0054] If the distance between the movement trajectory change of the moving part corresponding to the internal and external force wear type defect in the real-time fabric production line does not exceed the threshold of the movement trajectory change distance, and the amount of the transverse and longitudinal stretching span of the fabric corresponding to the internal wear type defect in the real-time fabric production line that exceeds the current fabric material toughness to meet the maximum stretching span value does not exceed the stretching span value excess threshold, then it is inferred that the probability of the corresponding type of defect is low, and the type other than the current type is regarded as the wear type, such as the surface deformation caused by the oxidation of the fabric material.

[0055] If the distance between the movement trajectory change of the moving part corresponding to the internal and external force wear type defect in the real-time fabric production line exceeds the threshold for the distance between movement trajectory changes, or if the amount by which the transverse and longitudinal stretching span of the fabric corresponding to the internal wear type defect in the real-time fabric production line exceeds the maximum stretching span value that the current fabric material toughness meets, it is inferred that the probability of the corresponding type of defect is high, the current type is identified as the wear type, and an alert is sent to the administrator.

[0056] The process for selecting the reference image is as follows:

[0057] After obtaining reference fabric images and the estimated causes of defects, the real-time production line is equipped with reference image samples to ensure the efficiency and clarity of the reference image sample acquisition, thereby improving the accuracy of image recognition of fabric defects.

[0058] Within the current production line's fluctuating production speed, the ratio between the rate of decrease in the number of missed frames per unit time during image acquisition by the current image acquisition device and the sharpness deviation value of adjacent frames is obtained. Only the ratio of the speed and sharpness values ​​is collected to infer the impact of the numerical fluctuations, without considering the issue of inconsistent units. The ratio between the rate of decrease in the number of missed frames per unit time during image acquisition by the current image acquisition device and the sharpness deviation value of adjacent frames is marked as production line impact data.

[0059] When the current image acquisition device acquires an image, it obtains the maximum reciprocating deviation value of the luminous flux received by the current light source in the corresponding fabric area of ​​the acquired image, and marks the maximum reciprocating deviation value of the luminous flux received by the current light source in the corresponding fabric area of ​​the acquired image as light source influence data;

[0060] And it is compared with the numerical ratio threshold and the maximum reciprocating deviation threshold respectively:

[0061] If the ratio of the rate of decrease in the number of missed frames per unit time and the sharpness deviation value of adjacent frames exceeds the threshold value, and the maximum reciprocating deviation value of the luminous flux received by the fabric area of ​​the corresponding acquired image does not exceed the maximum reciprocating deviation value threshold value, then it is inferred that the image acquisition of the current image acquisition device is qualified, and the image acquired in the current time period is used as a real-time reference image sample.

[0062] If the ratio between the rate of decrease in the number of missed frames per unit time and the sharpness deviation value of adjacent frames does not exceed the threshold value, or if the maximum reciprocating deviation value of the luminous flux received by the fabric area of ​​the corresponding acquired image exceeds the maximum reciprocating deviation value threshold value, then it is inferred that the image acquisition of the current image acquisition device is unqualified, and the image acquired in the current time period will not be used as a real-time reference image sample. In subsequent acquisitions, an additional acquisition device will be added or the current acquisition device will be adjusted.

[0063] The real-time defect comparison process is as follows:

[0064] After obtaining the real-time reference image sample, it is compared with the reference fabric image. The real-time reference image sample is extracted according to the process sequence of the reference fabric image to ensure that the production progress of the compared images is at the same process stage. The distance ratio of the overlapping distance to the non-overlapping distance of the fabric texture trajectory at the same location within the corresponding fabric area of ​​the reference fabric image and the extracted sample image is obtained. Simultaneously, the numerical deviation value of the image display brightness at any corresponding location between the reference fabric image and the extracted sample image is obtained. The distance ratio of the overlapping distance to the non-overlapping distance of the fabric texture trajectory at the same location within the corresponding fabric area of ​​the reference fabric image and the extracted sample image, and the numerical deviation value of the image display brightness at any corresponding location between the reference fabric image and the extracted sample image are marked as texture defect information and brightness defect information, respectively, and compared with distance ratio thresholds and brightness deviation thresholds, respectively.

[0065] If the ratio of the overlap distance to the non-overlap distance of the fabric texture trend trajectory at the same location within the fabric area corresponding to the reference fabric image and the sample extracted image exceeds the distance ratio threshold, and the numerical deviation of the image display brightness at any location corresponding to the reference fabric image and the sample extracted image does not exceed the brightness deviation threshold, then it is inferred that there is no defect risk in the current comparison, a processing defect-free signal is generated and sent to the administrator terminal.

[0066] If the ratio of the overlap distance to the non-overlap distance of the fabric texture trajectory at the same location within the corresponding fabric area of ​​the reference fabric image and the sample extracted image does not exceed the distance ratio threshold, or if the numerical deviation of the image display brightness at any location corresponding to the reference fabric image and the sample extracted image exceeds the brightness deviation threshold, it is inferred that there is a defect risk in the current comparison, a processing defect signal is generated, and the defect location is sent to the administrator terminal; after receiving the signal, the administrator terminal will rectify the current fabric processing production line.

[0067] The above formulas are all derived from software simulation using a large amount of data, and are selected to be close to the true values. The coefficients in the formulas are set by those skilled in the art based on the actual situation.

[0068] In use, this invention involves: image detection, acquiring images of historically processed fabrics, and inferring whether the currently processed fabric is suitable as a reference fabric for defect detection based on the acquired images; defect cause assessment, analyzing the real-time fabric processing environment and process to infer the cause of fabric defects during the current processing; image selection, simulating the processing steps based on the selected defect detection reference fabric, and acquiring images based on the simulated steps; and real-time defect comparison, performing image analysis and processing between the currently produced fabric and the defect detection reference fabric, and inferring whether the currently produced fabric has defects through image comparison.

[0069] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for intelligent detection of fabric defects based on image recognition analysis, characterized in that, The defect detection method steps are as follows: Based on the image inspection, images of historically processed fabrics are acquired, and the current historically processed fabrics are inferred from the acquired images to determine whether they are suitable as reference fabrics for defect detection. Inspection videos of the historical fabric processing process are acquired, and arbitrary segments of the inspection videos are extracted based on the historical fabric processing results to obtain the processing time periods of defect-free fabrics. Surface images of the fabrics within the corresponding segmented processing time periods are then collected, with the fabric width based on the fabric production width and the fabric length manually set by the number of images collected per unit time by the real-time production line image acquisition equipment to ensure that all acquired images meet the required clarity. The fabric surface is divided into several sub-regions, and a detection image of the fabric surface is constructed by selecting the number of sub-regions. Specifically, the pixel dimensions of each sub-region in the detection image are set to X×Y. Grayscale values ​​of the detection image are collected using a sensor, and H(x,y) is defined as the grayscale value at (x,y) within the sub-region of the detection image. The average grayscale value of the detection image is then obtained using a formula. The formula is: ; After obtaining the average gray value, the standard deviation of the gray value of the detected image is obtained using the formula: ; Defect cause assessment involves analyzing the real-time fabric processing environment and technology to infer the causes of fabric defects during the current processing. The defect cause assessment process is as follows: The system retrieves historical processing times for the current fabric production line and collects the types of fabric defects produced during those times, categorizing them into external wear and internal wear types. It also retrieves information on external force modifications and internal adjustment, comparing these with thresholds for changes in movement trajectory spacing and excess stretching span, respectively. If the external force modification information does not exceed the movement trajectory change spacing threshold, and the internal adjustment information does not exceed the stretch span value excess threshold, then the type other than the current type will be taken as the wear type. If the external force alteration information exceeds the movement trajectory change spacing threshold, or the internal adjustment information exceeds the stretching span value excess threshold, then the current type will be designated as the wear type. The image selection process involves simulating the processing steps of the selected defect detection reference fabric and generating images based on the simulated steps. The image selection process is as follows: During the current production line's fluctuating production rate based on actual production speed, production line impact data and light source impact data were obtained; and these were compared with the numerical ratio threshold and the maximum reciprocating deviation threshold, respectively. If the production line impact data exceeds the numerical ratio threshold, and the light source impact data does not exceed the maximum reciprocating deviation threshold, then the image collected in the current time period will be used as the real-time reference image sample; if the production line impact data does not exceed the numerical ratio threshold, or the light source impact data exceeds the maximum reciprocating deviation threshold, then the image collected in the current time period will not be used as the real-time reference image sample, and additional acquisition equipment will be added or the current acquisition equipment will be adjusted in subsequent acquisitions. Real-time defect comparison involves analyzing images of the fabric being produced in real-time against a reference fabric for defect detection. The comparison is used to infer whether any defects exist in the currently produced fabric. The real-time defect comparison process is as follows: The system compares the real-time reference image sample with the reference fabric image to obtain texture defect information and brightness defect information, and compares them with the distance ratio threshold and brightness deviation threshold respectively: if the texture defect information exceeds the distance ratio threshold and the brightness defect information does not exceed the brightness deviation threshold, it is inferred that the current comparison has no defect risk; if the texture defect information does not exceed the distance ratio threshold or the brightness defect information exceeds the brightness deviation threshold, it is inferred that the current comparison has a defect risk. If the average gray value and standard deviation of the corresponding sub-region are both within the corresponding set threshold range, the corresponding sub-region is marked as a defect-free sub-region; if either the average gray value or standard deviation of the corresponding sub-region is not within the corresponding set threshold range, the corresponding sub-region is marked as a defective sub-region; the surface of the collected fabric is segmented based on the coverage area of ​​the detected image as the segmentation parameter to obtain a real-time reference fabric, and a reference fabric image is obtained based on the real-time reference fabric; The external force modification information and the internal production adjustment information are respectively the distance of the change in the movement trajectory of the moving parts corresponding to the internal and external force wear type defects in the real-time production fabric line, and the amount by which the processing technology corresponding to the internal production wear type defects in the real-time production fabric line affects the transverse and longitudinal tensile span of the fabric and the current fabric material toughness meets the maximum tensile span value.

2. The intelligent detection method for fabric defects based on image recognition analysis according to claim 1, characterized in that, The production line impact data and light source impact data are respectively the ratio of the rate of decrease in the number of missed frames per unit time during image acquisition by the current image acquisition equipment to the sharpness deviation value of adjacent frames, and the maximum cyclic deviation value of the luminous flux received by the current light source in the fabric area of ​​the acquired image.

3. The intelligent detection method for fabric defects based on image recognition analysis according to claim 1, characterized in that, Texture defect information and brightness defect information are respectively the ratio of the overlapping distance to the non-overlapping distance of the fabric texture trend trajectory at the same position within the corresponding fabric area of ​​the reference fabric image and the real-time reference image sample extracted image, and the numerical deviation value of the image display brightness at any position of the reference fabric image and the real-time reference image sample extracted image.

Images