Textile fabric digital analysis method

Abstract

The application discloses a kind of textile fabric digital analysis method, steps include: collecting textile fabric front, back high-definition image, and carry out gray processing;Based on gray image, the ROI initial selected area of fabric texture is extracted;Obtain the copy of ROI initial selected area and determine the offset of copy relative to initial ROI area by template matching method, and then obtain the ROI revised selected area as the basic unit of textile fabric digital analysis;Marked image is obtained by using the trained deep learning yarn structure image recognition model to identify yarn structure and determine the type of textile fabric;The yarn structure on the marked image is mapped to the loom weaving diagram template corresponding to the fabric type, and the loom weaving schematic diagram of the fabric is obtained;The method has low operation difficulty, different types of fabric weaving process are analyzed by introducing traditional image processing and deep learning neural network, and the loom weaving schematic diagram can be directly obtained, the accuracy and work efficiency of fabric analysis link are improved.

Description

Technical Field

[0001] This invention relates to the field of textile fabric analysis and processing technology, and in particular to a method for analyzing textile fabric samples. Background Technology

[0002] Textile fabrics are an essential component of daily life and industrial applications, encompassing three main categories: woven fabrics, knitted fabrics, and nonwoven fabrics. Fabric structure is the inherent property of the fabric formed after yarn weaving; different structures give fabrics different styles and appearances. Currently, in the textile industry, analyzing fabric structure typically involves manually disassembling individual yarns, recording the interlacing or floating patterns between them, determining the weaving action of each yarn at that location based on its bending shape, and finally recording the analysis results. This analysis process generally takes 10-60 minutes depending on the complexity of the fabric, and suffers from low efficiency and significant susceptibility to worker experience and subjective biases, making it unsuitable for the high-quality, high-speed development demands of enterprises.

[0003] Image processing technology, characterized by its high processing speed and accuracy, has been widely applied in fields such as medicine, engineering, architecture, and other manufacturing industries with the continuous advancement and development of modern science and technology. Chinese patent application CN201511003862.0 proposes a method for recognizing woven fabric weave patterns, capable of identifying the three primary weave structures and their simple variations. Chinese patent application CN201811277402.0 relates to an analysis method and system for the weave structure of woven fabrics, which shows good recognition results for single-weave regions in fabric images, but recognition of other types of fabrics and composite weaves remains challenging. Therefore, there is an urgent need to develop a digital analysis method capable of rapidly acquiring woven patterns of various textile fabrics. Summary of the Invention

[0004] The purpose of this invention is to provide a method for digital analysis of textile fabrics that can perform digital analysis on different types of textile fabrics and obtain corresponding knitting patterns.

[0005] Therefore, the technical solution of the present invention is as follows:

[0006] A digital analysis method for textile fabrics, the specific implementation steps of which are as follows:

[0007] S1. Acquire high-resolution images of the front and back sides of the textile fabric in a planar unfolded state; wherein, the high-resolution image should contain at least three complete pattern cycle units, and the resolution of the fabric in the high-resolution image should be at least 0.1 mm.

[0008] S2. The high-definition image acquired in step S1 is converted to grayscale to obtain a grayscale image, and the initial ROI selection area of ​​the fabric texture in the image is extracted based on traditional image processing algorithms.

[0009] S3. Obtain the initial ROI selection area in the grayscale image to obtain a copy, and superimpose the copy onto the grayscale image obtained in step S2. Use template matching to perform translation, scaling, or rotation operations on the copy to determine the offset of the copy relative to the initial ROI area.

[0010] S4. Based on the offset in the length direction and the offset in the width direction obtained in step S3, the initial ROI selection area in the image obtained in step S2 is corrected to obtain the corrected ROI selection area; the corrected ROI selection area is cropped from the image as the basic unit for digital analysis of textile fabrics.

[0011] S5. Input the basic unit for digital analysis of textile fabrics into the trained deep learning yarn structure image recognition model to perform image recognition of the yarn structure of textile fabrics, obtain marked images with different yarn structures in the selection box and display the name of the yarn structure on the selection box, and obtain the type of textile fabric.

[0012] S6. Based on the textile fabric type obtained in step S5, retrieve the knitting pattern template corresponding to the textile fabric type; scale the knitting pattern template and the marked image obtained in step S5 to the same size, and then map the yarn structure knots identified in the marked image one by one onto the knitting pattern template to obtain the knitting pattern diagram of the fabric.

[0013] Among them, Figure 3(a) shows the knitting pattern template for weft-knitted fabrics; Figure 3(b) shows the knitting pattern template for warp-knitted fabrics; and Figure 3(c) shows the knitting pattern template for woven fabrics.

[0014] Further, in step S1, the high-definition image is acquired through an image acquisition system; the image acquisition system includes two cameras equipped with high-definition industrial lenses, a fabric clamp for holding the fabric, a light source, and a camera bracket; wherein,

[0015] The fabric clamping frame consists of a horizontal moving mechanism, a lifting moving mechanism, and a fabric clamping frame arranged sequentially from bottom to top. The bottom of the horizontal moving mechanism is fixed to the ground, the lifting moving mechanism is fixed to the slider of the horizontal moving mechanism, and the fabric clamping frame is vertically fixed to the top of the lifting moving mechanism. The fabric clamping frame moves back and forth along the length of the fabric clamping frame via the horizontal moving mechanism and along the height of the fabric clamping frame via the lifting moving mechanism. Two cameras equipped with high-definition industrial lenses are symmetrically arranged on both sides of the fabric clamping frame via camera brackets fixed to the ground. The lenses of each camera are horizontally facing the textile fabric, so that the high-definition images captured by the two cameras are the front and back images of the same area of ​​the textile fabric, respectively. Two light sources are symmetrically arranged on both sides of the fabric clamping frame, with each light source facing the textile fabric on the fabric clamping frame and set at a 45° downward angle.

[0016] Furthermore, the fabric clamping frame is a rectangular frame that is split into two parts, allowing the textile fabric to be clamped between the two split parts of the rectangular frame and fixed together by clamps, so that the textile fabric can be unfolded in a plane and hang freely to achieve image acquisition.

[0017] Furthermore, in step S3, the specific steps for determining the offset of the copy relative to the initial ROI region are as follows:

[0018] S301. Copy the initial ROI selection area from the image with the initial ROI selection area annotation box obtained in step S2 to obtain a copy of the ROI selection area;

[0019] S302. Place the copy on the initial ROI selection area bounding box of the image obtained in step S2. By translating the copy to the left or right along the length direction of the initial ROI selection area, find the adjacent position in the grayscale image that corresponds to and coincides with the texture of the copy. At this time, the offset of the copy in the length direction of the initial ROI selection area is the optimal length of the basic unit.

[0020] S303. Translate the copy upwards or downwards along the width direction of the initial ROI selection area. Find a position in the grayscale image that corresponds to and coincides with the texture of the copy. At this time, the offset of the copy in the width direction of the initial ROI selection area is the optimal width of the base unit.

[0021] Furthermore, the specific implementation steps of step S5 are as follows:

[0022] S501. Construct and train a deep learning model for yarn structure image recognition; among which...

[0023] (1) The deep learning yarn structure image recognition model is a model trained on a large number of different types of yarn structure samples based on a deep learning network. The deep learning network selected is Faster R-CNN, LeNet, AlexNet, ZFNet, VGGNet, GoogLeNet, ResNet, or YOLOV5.

[0024] (2) The training method for the deep learning yarn structure image recognition model is as follows:

[0025] Constructing yarn structure training samples: Two-dimensional yarn structure images of left open loop, right open loop, left closed loop, right closed loop, complete loop, tucked loop, left shifted loop, right shifted loop, non-woven loop, warp intersection, and weft intersection are obtained as input images; a rectangle is drawn outside each two-dimensional yarn structure image, and the yarn structure name is displayed on the rectangle to form a labeled image as the output image; there are no less than 2000 images of each type of two-dimensional yarn structure, thus obtaining a yarn structure training sample containing no less than 200,000 labeled images;

[0026] Model training: Using a two-dimensional yarn structure image as input information and a labeled image with a rectangle outside the yarn structure image and the name of the yarn structure displayed on the rectangle as output, the yarn structure training samples are sequentially input into the specified deep learning network to complete the training and optimization of the deep learning network and obtain a deep learning yarn structure image recognition model.

[0027] S502. Input the basic unit for digital analysis of textile fabric obtained in step S4 into the deep learning yarn structure image recognition model to obtain a marked image on the image with a rectangular frame set outside each yarn structure and the name of the yarn structure displayed on the rectangular frame.

[0028] S503. Determine the type of textile fabric based on the yarn structure name marked on the marked image, including: warp-knitted fabric, weft-knitted fabric, and woven fabric.

[0029] Compared with existing technologies, this digital analysis method for textile fabrics firstly uses conventional methods such as camera lenses to acquire images, which is not difficult to operate. Secondly, by introducing traditional image processing and deep learning neural network technologies, this method can analyze the weaving processes of different types of fabrics and output the corresponding results as the weaving pattern template for the corresponding textile fabric type, thereby improving the accuracy and efficiency of the fabric analysis process. Attached Figure Description

[0030] Figure 1 This is a flowchart of the digital analysis method for textile fabrics of the present invention;

[0031] Figure 2 This is a schematic diagram of the pattern acquisition system in step S1 of the digital analysis method for textile fabrics of the present invention;

[0032] Figure 3(a) is a template for the weft knitting pattern of a fabric.

[0033] Figure 3(b) is a template for the knitting pattern of warp-knitted fabrics;

[0034] Figure 3(c) is a template for the weaving pattern of woven fabric;

[0035] Figure 4(a) is a schematic diagram of the grayscale image A' obtained after the high-definition image A is processed by grayscale in step S2 in an embodiment of the present invention;

[0036] Figure 4(b) is a schematic diagram of the initial selection area of ​​fabric texture ROI extracted from grayscale image A' after step S2 in an embodiment of the present invention;

[0037] Figure 4(c) is a schematic diagram of the position corresponding to and coinciding with the texture of the copy found by the copy in the length direction and width direction respectively from the initial ROI selection direction of the grayscale image A' in the embodiment of the present invention after step S3;

[0038] Figure 4(d) is a schematic diagram of the ROI correction selection area obtained on the grayscale image A' after step S4 correction in an embodiment of the present invention;

[0039] Figure 4(e) is a schematic diagram of the grayscale image B' obtained after the high-definition image B is grayscaled in step S2 in an embodiment of the present invention;

[0040] Figure 4(f) is a schematic diagram of the initial selection area of ​​fabric texture ROI obtained from grayscale image B' after step S2 in an embodiment of the present invention;

[0041] Figure 4(g) is a schematic diagram of the position corresponding to and coinciding with the texture of the copy body found by the copy body in the length direction and width direction respectively from the initial ROI selection direction of the grayscale image B' in the embodiment of the present invention after step S3;

[0042] Figure 4(h) is a schematic diagram of the ROI correction selection area obtained on the grayscale image B' after step S4 correction in an embodiment of the present invention;

[0043] Figure 5(a) is a schematic diagram of the basic unit for digital analysis of the front side of the textile fabric cut out from Figure 4(d) in an embodiment of the present invention;

[0044] Figure 5(b) is a schematic diagram of the basic unit for digital analysis of the reverse side of textile fabric cut out from Figure 4(h) in an embodiment of the present invention;

[0045] Figure 5(c) shows an embodiment of the present invention.

[0046] Figure 5(d) shows an embodiment of the present invention.

[0047] Figure 5(e) shows an embodiment of the present invention. Detailed Implementation

[0048] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but the following embodiments are by no means intended to limit the present invention.

[0049] like Figure 1 As shown, the specific implementation steps of this digital analysis method for textile fabrics are as follows:

[0050] S1. Acquire high-resolution images of the front and back sides of the textile fabric in a planar unfolded state; wherein, the high-resolution image should contain at least three complete pattern cycle units, and the resolution of the fabric in the high-resolution image should be at least 0.1 mm.

[0051] See Figure 2 High-definition images are acquired through an image acquisition system; the image acquisition system includes two cameras 1 equipped with high-definition industrial lenses, a fabric clamp 2 for holding the fabric, two light sources 3, and a camera bracket 5; among which,

[0052] The fabric clamping frame 2 consists of a horizontal moving mechanism, a lifting moving mechanism, and a fabric clamping frame arranged sequentially from bottom to top. The bottom of the horizontal moving mechanism is fixed to the ground, the bottom of the lifting moving mechanism is fixed to the slider of the horizontal moving mechanism, and the fabric clamping frame is vertically fixed to the top of the lifting moving mechanism, allowing the fabric clamping frame to reciprocate along the length of the fabric clamping frame via the horizontal moving mechanism. The fabric clamping frame is a rectangular frame, which is divided into two parts and can be connected and fixed as one unit by clamps. When performing image acquisition, the textile fabric 4 can be clamped between the two parts of the rectangular frame and fixed as one unit by clamps, so that the textile fabric 4 can achieve image acquisition in a flat and freely hanging state. The fabric clamping frame 2 drives the textile fabric 4 to move up and down and / or left and right through its horizontal moving mechanism and lifting moving mechanism.

[0053] Two cameras 1 equipped with high-definition industrial lenses are symmetrically arranged on both sides of the fabric clamping frame 2 and are set at the same height by camera brackets 5 fixed to the ground. Each camera 1 is set with its lens horizontally facing the textile fabric 4 so that the high-definition images captured by the two cameras are the front and back images of the same area of ​​the textile fabric, respectively. Two light sources 3 are symmetrically arranged on both sides of the fabric clamping frame 2 and above the cameras 1 on the same side. Each light source is facing the textile fabric on the fabric clamping frame 2 and is set at a 45° downward angle.

[0054] In this embodiment, a high-definition image of the front side of the textile fabric (hereinafter referred to as Image A) and a high-definition image of the back side of the textile fabric (hereinafter referred to as Image B) are acquired using an image acquisition system.

[0055] S2. The high-definition image acquired in step S1 is converted to grayscale to obtain a grayscale image, and the initial ROI selection area of ​​the fabric texture in the grayscale image is extracted based on traditional image processing algorithms.

[0056] In this embodiment, Figure 4(a) shows the grayscale image A' obtained after image A has been grayscaled (see the pattern marked 9 in the figure), and Figure 4(e) shows the grayscale image B' obtained after image B has been grayscaled (see the pattern marked 13 in the figure); Figure 4(b) shows the image of the initial selection area of ​​the fabric texture ROI extracted from the grayscale image A' based on the relative position of pixels, fabric texture, and grayscale information (see the rectangle marked 10 in the figure); Figure 4(f) shows the image of the selection area of ​​the fabric texture ROI extracted from the grayscale image B' based on the relative position of pixels, fabric texture, and grayscale information (see the rectangle marked 14 in the figure); in Figures 4(b) and 4(f), the ROI selection area of ​​the fabric texture is specifically the image area within the white rectangle in the image; since the image processing algorithm for the initial selection area of ​​the fabric texture ROI is a mature technology, it will not be described in detail here;

[0057] S3. Obtain the initial ROI selection area in the grayscale image to obtain a copy, and superimpose the copy onto the grayscale image obtained in step S2. Use template matching to perform translation, scaling, or rotation operations on the copy to determine the offset of the copy relative to the initial ROI area.

[0058] The specific implementation steps of step S3 are as follows:

[0059] S301. Copy the initial ROI selection area from the image with the initial ROI selection area annotation box obtained in step S2 to obtain a copy of the ROI selection area;

[0060] S302. Place the copy on the initial ROI selection area bounding box of the image obtained in step S2. By translating the copy to the left or right along the length direction of the initial ROI selection area, find the adjacent position in the grayscale image that corresponds to and coincides with the texture of the copy. At this time, the offset of the copy in the length direction of the initial ROI selection area is the optimal length of the basic unit.

[0061] S303. Translate the copy upwards or downwards along the width direction of the initial ROI selection area. Find the adjacent position in the grayscale image that corresponds to and coincides with the texture of the copy. At this time, the offset of the copy in the width direction of the initial ROI selection area is the optimal width of the basic unit.

[0062] Figure 4(c) shows the positions of the copy that correspond to and coincide with the texture of the copy in the length and width directions, respectively, after steps S302 and S303, from the initial ROI selection direction of grayscale image A' (see the two rectangles marked 11 in the figure); Figure 4(g) shows the positions of the copy that correspond to and coincide with the texture of the copy in the length and width directions, respectively, after steps S302 and S303, from the initial ROI selection direction of grayscale image B' (see the two rectangles marked 15 in the figure).

[0063] S4. Based on the offset in the length direction and the offset in the width direction obtained in step S3, the initial ROI selection area in the image obtained in step S2 is corrected to obtain the corrected ROI selection area; the corrected ROI selection area is cropped from the image as the basic unit for digital analysis of textile fabrics.

[0064] Figure 4(d) shows the ROI correction selection area obtained on the grayscale image A' after correction in step S4 (see the rectangle marked 12 in the figure); Figure 4(h) shows the ROI correction selection area obtained on the grayscale image B' after correction in step S4 (see the rectangle marked 16 in the figure); Figure 5(a) shows a schematic diagram of the basic unit for digital analysis of the front side of the textile fabric cut out from Figure 4(d); Figure 5(b) shows a schematic diagram of the basic unit for digital analysis of the reverse side of the textile fabric cut out from Figure 4(h).

[0065] S5. Input the basic unit for digital analysis of textile fabrics into the trained deep learning yarn structure image recognition model to perform image recognition of the yarn structure of textile fabrics and obtain the type of textile fabric.

[0066] Specifically, the implementation steps of step S5 are as follows:

[0067] S501. Construct and train a deep learning model for yarn structure image recognition; among which...

[0068] (1) The deep learning yarn structure image recognition model is a model based on a deep learning network and trained with a large number of yarn structure samples of different types; among them, the deep learning network selected is Faster R-CNN network, LeNet network, AlexNet network, ZFNet network, VGGNet network, GoogLeNet network, ResNet, or YOLOV5 network.

[0069] (2) The training method for the deep learning yarn structure image recognition model is as follows:

[0070] 1) Constructing yarn structure training samples: Two-dimensional yarn structure images of left open loop, right open loop, left closed loop, right closed loop, complete loop, tucked loop, left shifted loop, right shifted loop, non-woven loop, warp intersection, and weft intersection are obtained as input images; a rectangle is drawn outside each two-dimensional yarn structure image and the yarn structure name is displayed on the rectangle to form a labeled image as the output image; there are no less than 2000 images of each type of two-dimensional yarn structure, thus obtaining a yarn structure training sample containing no less than 200,000 labeled images;

[0071] 2) Model training: Using a two-dimensional yarn structure image as input information and a labeled image with a rectangle outside the yarn structure image and the name of the yarn structure displayed on the rectangle as output, the yarn structure training samples are sequentially input into the specified deep learning network to complete the training and optimization of the deep learning network and obtain a deep learning yarn structure image recognition model.

[0072] S502. Input the basic unit for digital analysis of textile fabrics obtained in step S4 into the deep learning yarn structure image recognition model to obtain a marked image with different yarn structures selected and the name of the yarn structure displayed on the selection box.

[0073] S503. Determine the type of textile fabric based on the yarn structure name marked on the marked image; wherein, the type of fabric includes: warp-knitted fabric, weft-knitted fabric, and woven fabric;

[0074] In step S503, since different types of fabrics have different yarn structures, the yarn structure corresponds to the fabric type. Specifically, fabrics with features such as left open loops, right open loops, left closed loops, and right closed loops are warp-knitted fabrics; fabrics with complete loops, tucked loops, left shifted loops, right shifted loops, and non-knitted loops are weft-knitted fabrics; and fabrics with warp yarn intersections and weft yarn intersections are woven fabrics.

[0075] S6. Based on the textile fabric type obtained in step S5, retrieve the knitting pattern template corresponding to the textile fabric type; scale the knitting pattern template and the marked image obtained in step S5 to the same size, and then map the yarn structure knots identified in the marked image one by one onto the knitting pattern template to obtain the knitting pattern diagram of the fabric.

[0076] In this embodiment, the fabric can be determined to be a single-sided fabric at the initial stage of image acquisition. Then, after analysis in step S5, it is determined that the fabric belongs to weft knitting, and thus the textile fabric is determined to be a single-sided weft knitting fabric. Further, two weft knitting pattern templates as shown in 3(a) are retrieved respectively to correspond to the two marked images obtained in step S5, which correspond to the front and back of the textile fabric respectively. Taking the mapping of a weft knitting pattern template with the corresponding marked image on the front of the textile fabric as an example, the specific steps are as follows: 1) Scale the weft knitting pattern template and the marked image to the same size; 2) Process the marked image row by row in sequence, that is, mark the yarn structure identified in each row of the marked image in the corresponding part of each row of the weft knitting pattern according to the correspondence between the name and the code pattern.

[0077] Figure 5(c) shows a schematic diagram of the correspondence between yarn structure and code pattern. Referring to Figure 5(c), for single-sided fabrics, the yarn structure names displayed on the rectangles in the front and back marking images of the fabric are converted into corresponding codes and displayed in the lower half of each row of the online template image; for double-sided fabrics, the yarn structure names displayed on the front marking image selection box of the double-sided fabric are converted into corresponding codes and displayed in the lower half of each row of the online template image, and the yarn structure names displayed on the back marking image selection box of the double-sided fabric are converted into corresponding codes and displayed in the upper half of each row of the online template image.

[0078] Figure 5(d) shows the mapping result of Figure 5(a); Figure 5(e) shows the mapping result of Figure 5(b).

[0079] It should be noted that the parts of this invention not disclosed in detail belong to the well-known technology in the field; in addition, although the illustrative specific embodiments of this invention have been described above to enable those skilled in the art to understand this invention, it should be understood that this invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of this invention as defined and determined by the appended claims, and all inventions utilizing the concept of this invention are protected.

Claims

1. A digital analysis method for textile fabrics, characterized in that, The steps are as follows: S1. Acquire high-resolution images of the front and back sides of the textile fabric in a planar unfolded state; wherein, the high-resolution image should contain at least three complete pattern cycle units, and the resolution of the fabric in the high-resolution image should be at least 0.1 mm. S2. The high-definition image acquired in step S1 is converted to grayscale to obtain a grayscale image, and the initial ROI selection area of ​​the fabric texture in the image is extracted based on traditional image processing algorithms. S3. Obtain the initial ROI selection area in the grayscale image to obtain a copy, and superimpose the copy onto the grayscale image obtained in step S2. Use template matching to perform translation, scaling, or rotation operations on the copy to determine the offset of the copy relative to the initial ROI area. S4. Based on the offset in the length direction and the offset in the width direction obtained in step S3, the initial ROI selection area in the image obtained in step S2 is corrected to obtain the corrected ROI selection area; the corrected ROI selection area is cropped from the image as the basic unit for digital analysis of textile fabrics. S5. Input the basic unit for digital analysis of textile fabrics into the trained deep learning yarn structure image recognition model to perform image recognition of the yarn structure of textile fabrics, obtain marked images with different yarn structures in the selection box and display the name of the yarn structure on the selection box, and obtain the type of textile fabric. S6. Based on the textile fabric type obtained in step S5, retrieve the knitting pattern template corresponding to the textile fabric type; scale the knitting pattern template and the marked image obtained in step S5 to the same size, and then map the yarn structure knots identified in the marked image one by one onto the knitting pattern template to obtain the knitting pattern diagram of the fabric.

2. The digital analysis method for textile fabrics according to claim 1, characterized in that, In step S1, the high-definition image is acquired through an image acquisition system; the image acquisition system includes two cameras (1) equipped with high-definition industrial lenses, a fabric clamp (2) for holding the fabric, a light source (3), and a camera bracket (5); wherein, The fabric clamping frame (2) consists of a horizontal moving mechanism, a lifting moving mechanism, and a fabric clamping frame arranged sequentially from bottom to top. The bottom of the horizontal moving mechanism is fixed to the ground, the lifting moving mechanism is fixed to the slider of the horizontal moving mechanism, and the fabric clamping frame is vertically fixed to the top of the lifting moving mechanism. The fabric clamping frame moves back and forth along the length of the fabric clamping frame through the horizontal moving mechanism and moves back and forth along the height of the fabric clamping frame through the lifting moving mechanism. Two cameras (1) equipped with high-definition industrial lenses are symmetrically arranged on both sides of the fabric clamping frame (2) through camera brackets (5) fixed to the ground. The lens of each camera (1) is set horizontally towards the textile fabric (4) so ​​that the high-definition images captured by the two cameras are the front and back images of the same area of ​​the textile fabric, respectively. Two light sources (3) are symmetrically arranged on both sides of the fabric clamping frame (2), and each light source is set towards the textile fabric on the fabric clamping frame (2) at a 45° downward angle.

3. The digital analysis method for textile fabrics according to claim 2, characterized in that, The fabric clamping frame is a rectangular frame that is split into two parts, so that the textile fabric (4) can be clamped between the two parts of the rectangular frame and the three parts are fixed together by the clamps, so that the textile fabric (4) can be unfolded in a plane and hang freely to achieve image acquisition.

4. The digital analysis method for textile fabrics according to claim 1, characterized in that, In step S3, the specific steps for determining the offset of the copy relative to the initial ROI region are as follows: S301. Copy the initial ROI selection area from the image with the initial ROI selection area annotation box obtained in step S2 to obtain a copy of the ROI selection area; S302. Place the copy on the initial ROI selection area bounding box of the image obtained in step S2. By translating the copy to the left or right along the length direction of the initial ROI selection area, find the adjacent position in the grayscale image that corresponds to and coincides with the texture of the copy. At this time, the offset of the copy in the length direction of the initial ROI selection area is the optimal length of the basic unit. S303. Translate the copy upwards or downwards along the width direction of the initial ROI selection area. Find a position in the grayscale image that corresponds to and coincides with the texture of the copy. At this time, the offset of the copy in the width direction of the initial ROI selection area is the optimal width of the base unit.

5. The digital analysis method for textile fabrics according to claim 1, characterized in that, The specific implementation steps of step S5 are as follows: S501. Construct and train a deep learning model for yarn structure image recognition; among which... (1) The deep learning yarn structure image recognition model is a model trained on a large number of different types of yarn structure samples based on a deep learning network. The deep learning network selected is Faster R-CNN, LeNet, AlexNet, ZFNet, VGGNet, GoogLeNet, ResNet, or YOLOV5. (2) The training method for the deep learning yarn structure image recognition model is as follows: Constructing yarn structure training samples: Two-dimensional yarn structure images of left open loop, right open loop, left closed loop, right closed loop, complete loop, tucked loop, left shifted loop, right shifted loop, non-woven loop, warp intersection, and weft intersection are obtained as input images; a rectangle is drawn outside each two-dimensional yarn structure image, and the yarn structure name is displayed on the rectangle to form a labeled image as the output image; there are no less than 2000 images of each type of two-dimensional yarn structure, thus obtaining a yarn structure training sample containing no less than 200,000 labeled images; Model training: Using a two-dimensional yarn structure image as input information and a labeled image with a rectangle outside the yarn structure image and the name of the yarn structure displayed on the rectangle as output, the yarn structure training samples are sequentially input into the specified deep learning network to complete the training and optimization of the deep learning network and obtain a deep learning yarn structure image recognition model. S502. Input the basic unit for digital analysis of textile fabric obtained in step S4 into the deep learning yarn structure image recognition model to obtain a marked image on the image with a rectangular frame set outside each yarn structure and the name of the yarn structure displayed on the rectangular frame. S503. Determine the type of textile fabric based on the yarn structure name marked on the marked image, including: warp-knitted fabric, weft-knitted fabric, and woven fabric.

Images