Fabric-based visual inspection method and inspection system

Abstract

The application discloses a cloth-based visual detection method and a detection system, wherein the cloth-based visual detection method comprises the following steps: acquiring a cloth image, wherein the cloth image is obtained by shooting the cloth during movement; performing image processing on the cloth image, and displaying a plurality of defect areas of the cloth image in a stripe-shaped area; positioning the distribution of the plurality of defect areas, determining the pulling direction of the defect area based on the distribution of the plurality of defect areas; taking the pulling direction of the defect area as a first reference factor, taking the pattern of the defect area as a second reference factor, determining the defect type of the defect area based on the first reference factor and the second reference factor; segmenting the defect area, acquiring features of the segmented graph, and classifying the features based on a neural network; constructing a defect learning model based on previous data, constructing a framework of the defect learning model in a progressive algorithm, and combining the neural network and the defect learning model to form a self-adaptive enhancement algorithm.

Description

Technical Field

[0001] This invention relates to the field of visual inspection technology, and more particularly to a visual inspection method and system based on fabric. Background Technology

[0002] With the development of technology, fabrics are used in various clothing fields. During the production process, fabrics are stretched, and defects are easily formed during the stretching process. In the current technology, the type of fabric defect is mainly determined by the shape of the defect area. However, the shape of the defect area can appear differently at different angles, resulting in low accuracy of the current fabric inspection. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a visual inspection method and system based on fabric. The defect type of the defect area is determined by a first reference factor and a second reference factor. The defect type fully considers the stretching direction and defects of the fabric. Furthermore, an adaptive enhancement algorithm is formed by combining a neural network and a defect learning model. This adaptive enhancement algorithm uses the external environment as a regular parameter and outputs the defect level corresponding to the defect type to further determine the defect level. This allows for accurate identification of the severity of each defect in the fabric, avoiding human observation and misidentification, and greatly improving the accuracy of fabric inspection.

[0004] To address the aforementioned technical problems, this invention provides a visual inspection method based on fabric, comprising: acquiring a fabric image, wherein the fabric image is obtained by photographing fabric during movement; performing image processing on the fabric image and displaying multiple defect regions of the fabric image as striped areas; locating the distribution of the multiple defect regions and determining the pulling direction of the defect regions based on the distribution of the multiple defect regions; using the pulling direction of the defect regions as a first reference factor and the pattern of the defect regions as a second reference factor, determining the defect type of the defect regions based on the first reference factor and the second reference factor; segmenting the defect regions and acquiring features from the segmented images for feature classification based on a neural network; constructing a defect learning model based on previous data and constructing the framework of the defect learning model using a progressive algorithm; combining the neural network and the defect learning model to form an adaptive enhancement algorithm, which uses the external environment as a regular parameter and outputs the defect level corresponding to the defect type.

[0005] In addition, this invention also provides a fabric-based visual inspection system, comprising: an acquisition module for acquiring fabric images, wherein the fabric images are obtained by photographing the fabric during movement; a display module for performing image processing on the fabric images and displaying multiple defect areas of the fabric images as striped regions; a positioning module for locating the distribution of the multiple defect areas and determining the pulling direction of the defect areas based on the distribution of the multiple defect areas; a type module for using the pulling direction of the defect areas as a first reference factor and the pattern of the defect areas as a second reference factor, and determining the defect type of the defect areas based on the first reference factor and the second reference factor; a classification module for segmenting the defect areas and acquiring features from the segmented graphics for feature classification based on a neural network; and a learning module for constructing a defect learning model based on previous data, constructing the framework of the defect learning model using a progressive algorithm, and combining the neural network and the defect learning model to form an adaptive enhancement algorithm, which uses the external environment as a regular parameter and outputs the defect level corresponding to the defect type.

[0006] In this embodiment of the invention, the method displays multiple defect areas of the fabric image as striped regions, highlighting the patterns of each defect area, with the patterns serving as a second reference factor. The pulling direction of the defect areas is determined based on their distribution, and this pulling direction is used as a first reference factor. Then, the defect type of the defect area is determined using both the first and second reference factors. This defect type fully considers the pulling direction and defects of the fabric, and combines a neural network and a defect learning model to form an adaptive enhancement algorithm. This adaptive enhancement algorithm uses the external environment as a regular parameter and outputs the defect level corresponding to the defect type to further determine the defect level, thereby accurately identifying the severity of each defect in the fabric, avoiding human observation and misidentification, and greatly improving the accuracy of fabric detection. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 This is a flowchart illustrating the fabric-based visual inspection method in an embodiment of the present invention. Figure 2This is a schematic diagram of the structural composition of the fabric-based visual inspection system in an embodiment of the present invention; Figure 3 This is a hardware diagram of an electronic device according to an exemplary embodiment. Detailed Implementation

[0009] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0010] Example Please see Figure 1 , Figure 1 This is a flowchart illustrating the fabric-based visual inspection method in an embodiment of the present invention.

[0011] like Figure 1 As shown, a visual detection method based on fabric includes: S11: Acquire a fabric image, wherein the fabric image is obtained based on a photograph of the fabric during its movement; In a specific implementation of this invention, the specific steps can be as follows: during the movement of the fabric, a visual detector captures the dynamically moving fabric and forms video of each dynamic frame; based on the splicing of the video of each dynamic frame, multiple regional images are obtained in the unfolded state; the multiple regional images are sorted according to time periods, and the port boundaries of each regional image are detected; the docking position is determined based on the trajectory points of the port boundaries, and the trajectory points of two adjacent regional images are compared; if the trajectory points of two adjacent regional images are in the same position, the two adjacent regional images are spliced ​​together, and the fabric image is gradually formed.

[0012] In order to adapt to shooting the fabric in both static and dynamic environments, the visual detector captures the dynamically moving fabric and forms video of each dynamic frame. In the unfolded state, multiple region images are obtained to facilitate the visualization of each dynamic frame video. In addition, the docking position is determined based on the trajectory points of the port boundary, and the trajectory points of two adjacent region images are compared. If the trajectory points of two adjacent region images are in the same position, the two adjacent region images are stitched together to gradually form the fabric image. The stitching of adjacent region images improves the ease of forming the fabric image and does not limit the number of region images, thereby overcoming the problem of excessive fabric length.

[0013] S12: Perform image processing on the fabric image and display multiple defect areas of the fabric image as striped areas; In a specific implementation of this invention, the specific steps may be as follows: perform grayscale processing on the fabric image to form a grayscale image; dynamically adjust the grayscale levels of the grayscale image and enhance the display degree of defective areas in the grayscale image; when the display degree reaches a preset display degree, segment the defective area and fill it with stripes along the texture of the defective area to form a striped area; display multiple defective areas of the fabric image as striped areas; perform individual noise reduction on the defective areas, and select mean filtering, median filtering, or neighborhood averaging according to the degree of damage to the defective areas to obtain the repaired defective areas.

[0014] Specifically, the grayscale levels of the grayscale image are dynamically adjusted to enhance the display of defective regions in the grayscale image. The grayscale levels of the grayscale image are adjusted dynamically to adjust the display of defective regions. Only defective regions that meet the requirements are segmented to ensure the accuracy of subsequent image segmentation.

[0015] Furthermore, multiple defect areas of the fabric image are displayed as striped regions. Denoising is performed individually on each defect area, and mean filtering, median filtering, or neighborhood averaging is selectively chosen based on the degree of damage to obtain the corrected defect areas. Since the advantages of mean filtering, median filtering, and neighborhood averaging differ, and different defect areas are matched to their respective advantages, targeted correction is beneficial. For example, mean filtering is chosen when the degree of damage to the defect area is low, median filtering is chosen when the degree of damage is medium, and neighborhood averaging is chosen when the degree of damage is high. Multiple methods can also be combined for specific scenarios.

[0016] At this point, the mean filter is a relatively intuitive filtering method. It employs the idea of ​​template computation, which is essentially a neighborhood operation. The median filter is a common nonlinear spatial filtering method. Its basic principle is to replace the value of a point in a digital image or sequence with the median value of all points in its neighborhood. Unlike the mean filter, the median filter selects a template centered on the target pixel, first statistically sorting all pixel values ​​within the template, and then selecting the median value to replace the target pixel value. The number of pixels in the template is generally odd, and the template shape is typically circular, square, or cross-shaped. The neighborhood averaging method adds the gray value of a pixel in the original image to the gray values ​​of its neighboring pixels, and then uses the average of the sum as the gray value of that pixel.

[0017] S13: Locate the distribution of the multiple defective regions, and determine the pulling direction of the defective regions based on the distribution of the multiple defective regions; In a specific implementation of this invention, the specific steps may be as follows: establish a coordinate origin based on the lower left corner of the fabric image, establish a fabric coordinate map, and project multiple defective regions onto the fabric coordinate map; determine the center of the defective region based on the polygonal distance of the defective region, and use the center of the defective region as the positioning coordinate of the defective region; determine the distance between multiple defective regions based on the center-to-center distance of the centers of multiple defective regions, and introduce the stretching direction of the fabric as the direction vector of the distance; input the distance and the direction vector into the stretching direction learning model, and the stretching direction learning model outputs the stretching direction of the defective region.

[0018] Specifically, the defective area is projected based on the fabric coordinate map, and the distance between multiple defective areas is determined based on the center-to-center distance of the centers of multiple defective areas. In addition, the distance and the direction vector are input into the tensile direction learning model, and the tensile direction learning model outputs the tensile direction of the defective area. The tensile direction of the defective area can be determined from multiple dimensions to ensure the detection accuracy of the tensile direction of the defective area.

[0019] S14: Using the pulling direction of the defective area as a first reference factor and the pattern of the defective area as a second reference factor, determine the defect type of the defective area based on the first reference factor and the second reference factor; In a specific implementation of this invention, the specific steps may be as follows: extracting an image of the defective region based on the stripe direction of the defective region, and determining the pattern of the defective region based on the image; using the stretching direction of the defective region as a first reference factor, and configuring a first weight ratio for the first reference factor; using the pattern of the defective region as a second reference factor, and configuring a second weight ratio for the second reference factor, wherein the second weight ratio and the first weight ratio have a multiple integral relationship; determining the defect coefficient of the defective region based on the first reference factor and the second reference factor, wherein the defect coefficient can be changed by adding human data; if the defect coefficient does not reach a preset coefficient threshold, triggering human adjustment, and determining the final defect coefficient by increasing the human coefficient; determining the defect type of the defective region based on the final defect coefficient.

[0020] The first reference factor and the second reference factor are assigned corresponding weight ratios, and the second weight ratio and the first weight ratio have a multiple integral relationship, so as to construct a matching relationship between the first reference factor and the second reference factor. Based on the first reference factor and the second reference factor, the defect coefficient of the defect area is determined, ensuring the accuracy of the defect coefficient of the defect area, and thus realizing the determination of the defect type of the defect area.

[0021] In addition, if the defect coefficient does not reach the preset threshold, manual adjustment is triggered, and the final defect coefficient is determined by increasing the manual coefficient, ensuring a certain degree of human freedom in the defect coefficient, which can cope with different scenarios.

[0022] S15: Segment the defective region and extract features from the segmented image to perform feature classification based on a neural network; In the specific implementation of this invention, the specific steps include: dividing the defective region into equal-distance segments extending outward from the center and towards the edge of the defective region to form a segmented pattern; adjusting the placement of the segmented pattern according to the defect orientation of the segmented pattern, and ensuring that the placement of the segmented pattern is parallel to the defect orientation of the segmented pattern; individually obtaining the placement starting point of the placed segmented pattern, and performing feature acquisition along the placement starting point to determine sub-features; using the sub-features as elements, inputting the elements into a neural network, and performing feature classification under the autonomous allocation of the neural network, wherein the neural network is a binary neural network.

[0023] In this process, the segmentation is based on equal-distance division from the center outwards, extending to the edge of the defect area to form a segmented pattern. At this time, all segmented patterns are equidistant and have equal areas, but different outer contours. To adapt to different outer contours, the placement of the segmented patterns is adjusted according to the defect orientation of the segmented patterns to ensure the uniformity of the graphic orientation of the determined sub-features, reduce the influence of orientation, and achieve accurate acquisition at the same angle.

[0024] The sub-features are used as elements and input into a neural network. Feature classification is performed under the autonomous allocation of the neural network. The elements are automatically allocated based on the neural network to facilitate the classification and processing using the similar computing capabilities of the neural network, thereby improving the neural network's ability and efficiency in allocating elements.

[0025] S16: Construct a defect learning model based on previous data, and construct the framework of the defect learning model using a progressive algorithm. Combine the neural network and the defect learning model to form an adaptive enhancement algorithm. This adaptive enhancement algorithm takes the external environment as a regular parameter and outputs the defect level corresponding to the defect type.

[0026] In the specific implementation of this invention, the specific steps include: acquiring past data and classifying the results based on the past data; performing same-category learning on the classified data packets and constructing a defect learning model, wherein the defect learning model contains at least 100,000 data packets; introducing a preset correction model and embedding the preset correction model into the defect learning model, and triggering the defect learning model when the output result of the defect learning model does not meet the preset precision; the defect learning model corrects the results that do not meet the preset precision and performs corresponding correction based on the defect type; acquiring the matching coefficient of the neural network and the matching coefficient of the defect learning model; if the difference between the matching coefficient of the neural network and the matching coefficient of the defect learning model is within a preset range, combining the neural network and the defect learning model, wherein the neural network and the defect learning model are embedded as sub-elements into the adaptive enhancement algorithm; performing integral processing on the neural network and the defect learning model respectively based on the adaptive enhancement algorithm; embedding the network algorithm of the neural network into the learning algorithm of the defect learning model, and forming the adaptive enhancement algorithm with the learning algorithm as the main component.

[0027] Among them, the defect learning model is learned based on at least 100,000 data packets, and the learning and model building are based on multiple different results, which greatly improves the accuracy of the defect learning model and enables it to handle problems in different scenarios. At this time, the results corresponding to previous data are classified; and the same category learning is performed on the classified data packets.

[0028] A preset correction model is introduced and embedded into the defect learning model. When the output of the defect learning model does not meet the preset accuracy, the defect learning model is triggered. The defect learning model corrects the results that do not meet the preset accuracy, thereby further modifying the results and ensuring the accuracy of the final results. Furthermore, the means of triggering the preset correction model in the later stages are used to monitor the results in a timely manner.

[0029] In addition, the matching coefficients of the neural network and the defect learning model are obtained; if the difference between the matching coefficients of the neural network and the defect learning model is within a preset range, the neural network and the defect learning model are combined, wherein the neural network and the defect learning model are embedded as sub-elements into the adaptive enhancement algorithm; the neural network and the defect learning model are respectively integrated based on the adaptive enhancement algorithm; the network algorithm of the neural network is embedded into the learning algorithm of the defect learning model, and the adaptive enhancement algorithm with the learning algorithm as the main component is formed.

[0030] In this embodiment of the invention, the defect type of the defect area is determined by the method of the present invention using a first reference factor and a second reference factor. The defect type fully considers the stretching direction and defects of the fabric, and combines a neural network and a defect learning model to form an adaptive enhancement algorithm. The adaptive enhancement algorithm uses the external environment as a regular parameter and outputs the defect level corresponding to the defect type to further determine the level of the defect, thereby accurately identifying the severity of each defect in the fabric, avoiding human observation and misidentification, and greatly improving the accuracy of fabric detection.

[0031] Example Please see Figure 2 , Figure 2 This is a schematic diagram of the structural composition of a fabric-based visual inspection system according to an embodiment of the present invention.

[0032] like Figure 2 As shown, a fabric-based visual inspection system includes: Acquisition module 21: used to acquire fabric images, wherein the fabric images are obtained based on photos of the fabric during movement; Display module 22: used to perform image processing on the fabric image and display multiple defect areas of the fabric image as striped areas; Positioning module 23: used to locate the distribution of the multiple defective regions and determine the pulling direction of the defective regions based on the distribution of the multiple defective regions; Type module 24: used to determine the defect type of the defect area based on the first reference factor and the second reference factor, taking the pulling direction of the defect area as the first reference factor and the pattern of the defect area as the second reference factor; Classification module 25: used to segment the defective region and acquire features from the segmented image to perform feature classification based on a neural network; Learning Module 26: Used to build a defect learning model based on previous data, and to construct the framework of the defect learning model using a progressive algorithm. It combines the neural network and the defect learning model to form an adaptive enhancement algorithm. The adaptive enhancement algorithm takes the external environment as a regular parameter and outputs the defect level corresponding to the defect type.

[0033] This invention provides a visual inspection method and system for fabric. The method determines the defect type of a defect area using a first reference factor and a second reference factor. This defect type fully considers the fabric's stretching direction and defects. Furthermore, it combines a neural network and a defect learning model to form an adaptive enhancement algorithm. This algorithm uses the external environment as a regular parameter and outputs the defect level corresponding to the defect type to further determine the defect level. This accurately identifies the severity of each defect in the fabric, avoiding human observation and misidentification, and greatly improving the accuracy of fabric inspection.

[0034] Example Please see Figure 3 See below for reference. Figure 3 To describe an electronic device 40 according to this embodiment of the present invention. Figure 3 The electronic device 40 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of the present invention.

[0035] like Figure 3 As shown, the electronic device 40 is manifested in the form of a general-purpose computing device. The components of the electronic device 40 may include, but are not limited to: at least one processing unit 41, at least one storage unit 42, and a bus 43 connecting different system components (including storage unit 42 and processing unit 41).

[0036] The storage unit stores program code that can be executed by the processing unit 41, causing the processing unit 41 to perform the steps described in the "Embodiment Methods" section of this specification according to various exemplary embodiments of the present invention.

[0037] Storage unit 42 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 421 and / or a cache memory unit 422, and may further include a read-only memory unit (ROM) 423.

[0038] Storage unit 42 may also include a program / utility 424 having a set (at least one) program module 425, such program module 425 including but not limited to: operating system, one or more application programs, other program modules and program data, each of these examples or some combination of these may include an implementation of a network environment.

[0039] Bus 43 can represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the multiple bus structures.

[0040] Electronic device 40 can also communicate with one or more external devices (e.g., keyboard, pointing device, Bluetooth device, etc.), and with one or more devices that enable a user to interact with electronic device 40, and / or with any device that enables electronic device 40 to communicate with one or more other computing devices (e.g., router, modem, etc.). This communication can be performed via input / output (I / O) interface 45. Furthermore, electronic device 40 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 46. Figure 3 As shown, network adapter 46 communicates with other modules of electronic device 40 via bus 43. It should be understood that, although... Figure 3 As not shown, other hardware and / or software modules may be used in conjunction with electronic device 40, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0041] From the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, terminal device, or network device, etc.) to execute the methods according to the embodiments of this disclosure.

[0042] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. This program can be stored in a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), a magnetic disk, or an optical disk, etc. Furthermore, it stores computer program instructions, which, when executed by a computer, cause the computer to perform the methods described above.

[0043] Furthermore, the above description of the fabric-based visual inspection method and system provided by the embodiments of the present invention has been detailed. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A visual inspection method based on fabric, characterized in that, include: Acquire images of the fabric, wherein the images are captured based on images of the fabric during its movement; The fabric image is processed, and multiple defect areas of the fabric image are displayed as striped regions, which are formed by filling the stripes along the texture of the defect areas. The distribution of the multiple defective regions is located, and the pulling direction of the defective regions is determined based on the distribution of the multiple defective regions; Using the stretching direction of the defective region as a first reference factor and the pattern of the defective region as a second reference factor, the defect type of the defective region is determined based on the first and second reference factors. This includes: extracting an image of the defective region based on the stripe direction of the defective region, and determining the pattern of the defective region based on the image; using the stretching direction of the defective region as the first reference factor and assigning a first weight ratio to the first reference factor; using the pattern of the defective region as the second reference factor and assigning a second weight ratio to the second reference factor, wherein the second weight ratio and the first weight ratio have a multiple integral relationship; determining the defect coefficient of the defective region based on the first and second reference factors, where the defect coefficient changes with the increase of human-generated data; if the defect coefficient does not reach a preset coefficient threshold, human intervention is triggered, and a final defect coefficient is determined with the increase of the human-generated coefficient; and determining the defect type of the defective region based on the final defect coefficient. The defective region is segmented, and features are extracted from the segmented images to perform feature classification based on a neural network. A defect learning model is constructed based on previous data, and a framework for the defect learning model is constructed using a progressive algorithm. The neural network and the defect learning model are combined to form an adaptive enhancement algorithm. This adaptive enhancement algorithm takes the external environment as a regular parameter and outputs the defect level corresponding to the defect type.

2. The fabric-based visual inspection method according to claim 1, characterized in that, The acquisition of fabric images, wherein the fabric images are obtained based on images of the fabric during movement, includes: During the movement of the fabric, a visual detector captures the dynamically moving fabric and generates video of each dynamic frame. Based on the splicing of videos from each dynamic frame, and in the unfolded state, multiple region images are obtained; The multiple region images are sorted according to time periods, and the port boundaries of each region image are detected. The docking position is determined based on the trajectory points of the port boundary, and the trajectory points of two adjacent region images are compared. If the trajectory points of two adjacent region images are in the same position, the two adjacent region images are stitched together to gradually form the fabric image.

3. The fabric-based visual inspection method according to claim 2, characterized in that, The step of image processing the fabric image and displaying multiple defect areas of the fabric image as striped regions includes: The fabric image is processed to grayscale to form a grayscale image; The grayscale levels of the grayscale image are dynamically adjusted to enhance the visibility of defective areas in the grayscale image. When the display level reaches the preset display level, the defect area is segmented, and stripes are filled along the texture of the defect area to form a striped area. Multiple defect areas in the fabric image are displayed as striped regions; Denoising is performed separately on the defective region, and mean filtering, median filtering, or neighborhood averaging are selected one by one according to the degree of damage to the defective region to obtain the repaired defective region.

4. The fabric-based visual inspection method according to claim 3, characterized in that, The step of locating the distribution of the multiple defective regions and determining the pulling direction of the defective regions based on their distribution includes: A coordinate origin is established based on the lower left corner of the fabric image, a fabric coordinate map is created, and multiple defective areas are projected onto the fabric coordinate map. The edge of the defective region is used to determine the center of the defective region, and the center of the defective region is used as the positioning coordinates of the defective region. The distance between the multiple defective regions is determined based on the center-to-center distance between the centers of the multiple defective regions, and the stretching direction of the fabric is introduced as the direction vector of the distance. The distance and the direction vector are input into the tension direction learning model, and the tension direction learning model outputs the tension direction of the defect area.

5. The fabric-based visual inspection method according to claim 1, characterized in that, The step of segmenting the defective region and extracting features from the segmented image for feature classification based on a neural network includes: The defective region is divided into segments, with equal distances extending outward from the center and towards the edge of the defective region, to form a segmented pattern. The placement of the segmented graphic is adjusted according to the orientation of the defects in the segmented graphic, and the placement of the segmented graphic is parallel to the orientation of the defects in the segmented graphic. The starting point of the segmented graphic after placement is obtained separately, and the features of the segmented graphic are obtained along the starting point to determine the sub-features; The sub-features are used as elements, and these elements are input into a neural network, where feature classification is performed under the autonomous allocation of the neural network.

6. The fabric-based visual inspection method according to claim 5, characterized in that, The defect learning model is constructed based on previous data, and a framework for the defect learning model is built using a progressive algorithm. This framework, combined with the neural network and the defect learning model, forms an adaptive enhancement algorithm. This adaptive enhancement algorithm uses the external environment as a regular parameter and outputs the defect level corresponding to the defect type, including: Obtain historical data and categorize the results based on that data. For the categorized data packets, perform same-category learning and build a defect learning model. At this point, the defect learning model contains at least 100,000 data packets. A preset correction model is introduced and embedded into the defect learning model. When the result output by the defect learning model does not meet the preset accuracy, the defect learning model is triggered. The defect learning model corrects results that do not meet the preset accuracy and makes corresponding corrections based on the defect type.

7. The fabric-based visual inspection method according to claim 6, characterized in that, The framework for constructing a defect learning model based on past data and using a progressive algorithm to build the defect learning model, combined with the neural network and the defect learning model to form an adaptive enhancement algorithm, which uses the external environment as a regular parameter and outputs the defect level corresponding to the defect type, also includes: Obtain the matching coefficients of the neural network and the defect learning model; If the difference between the matching coefficient of the neural network and the matching coefficient of the defect learning model is within a preset range, then the neural network and the defect learning model are combined, wherein the neural network and the defect learning model are embedded as sub-elements into the adaptive enhancement algorithm. The adaptive enhancement algorithm is used to perform integral processing on the neural network and the defect learning model respectively; The network algorithm of the neural network is embedded into the learning algorithm of the defect learning model to form the adaptive enhancement algorithm with the learning algorithm as the main component.

8. A fabric-based visual inspection system, characterized in that, The fabric-based visual inspection system includes: Acquisition module: used to acquire fabric images, wherein the fabric images are obtained based on photos of the fabric during movement; Display module: used to perform image processing on the fabric image and display multiple defect areas of the fabric image as striped areas, the striped areas being formed by filling stripes along the texture of the defect areas; Positioning module: used to locate the distribution of multiple defective regions and determine the pulling direction of the defective regions based on the distribution of the multiple defective regions; Type Module: Used to determine the defect type of the defect area based on the stretching direction of the defect area as a first reference factor and the pattern of the defect area as a second reference factor, including: extracting an image of the defect area based on the stripe direction of the defect area, and determining the pattern of the defect area based on the image; using the stretching direction of the defect area as the first reference factor and configuring a first weight ratio for the first reference factor; using the pattern of the defect area as the second reference factor and configuring a second weight ratio for the second reference factor, wherein the second weight ratio and the first weight ratio have a multiple integral relationship; determining the defect coefficient of the defect area based on the first reference factor and the second reference factor, wherein the defect coefficient changes with the increase of human data; if the defect coefficient does not reach a preset coefficient threshold, human adjustment is triggered, and a final defect coefficient is determined with the increase of the human coefficient; and determining the defect type of the defect area based on the final defect coefficient. Classification module: used to segment the defective region and extract features from the segmented image for feature classification based on a neural network; Learning module: It is used to build a defect learning model based on previous data and construct the framework of the defect learning model with a progressive algorithm. It combines the neural network and the defect learning model to form an adaptive enhancement algorithm. The adaptive enhancement algorithm takes the external environment as a regular parameter and outputs the defect level corresponding to the defect type.

Images