Digital fabric physical attribute identification method and system based on pattern identification technology

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a digital fabric physical attribute recognition method and system based on a pattern recognition technology, and the method comprises the steps: constructing a digital fabric picture library, determining the influence significance of each physical attribute parameter on the appearance of a fabric, performing corresponding sampling on each physical attribute parameter based on the influence significance to obtain a sample physical attribute of the digital fabric picture; determining an image area of the digital fabric picture, and endowing image attributes of the image area to the digital fabric picture; constructing a training sample based on the image attribute and the sample physical attribute of the digital fabric picture, and inputting the training sample into a neural network for training to obtain a physical attribute recognition model of the digital fabric; performing image data acquisition on the real fabric, preprocessing the acquired image data, and inputting the preprocessed image data into the physical attribute recognition model to obtain predicted physical attributes of the real fabric; according to the invention, the physical attribute parameters of the fabric can be accurately obtained.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method and system for identifying physical properties of digital fabrics based on graphic recognition technology. Background Art

[0002] Digital fabric is a technology that uses digital technology to simulate the characteristics of real fabrics. It collects basic scientific data of real fabrics and combines various targeted algorithms of computer software, and uses high-performance computing clusters for simulation to achieve high-precision virtual simulation of fabrics. It can simulate the texture properties of fabrics through various fabric image information such as texture maps, normal maps, displacement maps, etc. It can also simulate the physical properties of fabrics by setting various fabric data information such as yarn strength, bending strength, density, thickness, etc.

[0003] The simulation authenticity of digital fabrics depends largely on the accurate capture and reproduction of physical property information. Virtual simulation of mechanical properties simulates the physical properties of real fabrics, such as elasticity, thickness, weight, drape, etc., and simulates the behavior of fabrics under stretching, bending and compression, ensuring that the mechanical response of virtual fabrics is consistent with that of real fabrics, giving virtual fabrics realistic physical properties.

[0004] In current technology, direct measurement with professional measuring instruments is usually used to obtain fabric physical properties: professional fabric physical property measuring instruments, such as Style3D's tensile measuring instrument Style3D SST1000 and bending strength measuring instrument Style3D SBE1000, are used to accurately test the physical properties of fabrics. The measured data can be directly applied to Style3D software to achieve digitalization and virtual simulation of fabrics. Similarly, these data can also be manually input into other virtual simulation software to build digital fabrics with corresponding physical properties.

[0005] However, this traditional method is time-consuming and costly when processing a large number of fabric samples, and it is difficult to meet the requirements of high-efficiency and high-precision recognition. Summary of the invention

[0006] In order to meet the needs of users for a more personalized and interactive multimedia experience, the present invention proposes a digital fabric physical property recognition method and system based on graphic recognition technology, which can accurately obtain the physical property parameters of the fabric without the need for traditional measuring tools.

[0007] In order to achieve the above object, the present invention provides the following technical solutions:

[0008] On the one hand, an embodiment of the present invention provides a method for identifying physical properties of digital fabrics based on pattern recognition technology, the method comprising the following steps:

[0009] S100, constructing a digital fabric image library, wherein the digital fabric image library includes a plurality of digital fabric images, each of which is configured with corresponding physical property parameters;

[0010] S200, determining the significance of the influence of each physical property parameter on the appearance of the fabric, and sampling each physical property parameter accordingly based on the significance of the influence to obtain sample physical properties of the digital fabric image;

[0011] S300, determining an image region of the digital fabric image, and assigning the area, graphic features and inflection points of the image region as image attributes to the digital fabric image;

[0012] S400, constructing a training sample based on the image attributes and sample physical attributes of the digital fabric image, inputting the training sample into a neural network for training, and obtaining a physical attribute recognition model of the digital fabric; wherein the input data of the training sample is the image attribute of the digital fabric image, and the output data of the training sample is the sample physical attribute of the digital fabric image;

[0013] S500, collecting image data of real fabrics, inputting the collected image data into a physical property recognition model, and obtaining predicted physical properties of the real fabrics.

[0014] Optionally, in S200, determining the significance of the influence of each physical property parameter on the fabric appearance, sampling each physical property parameter accordingly based on the significance of the influence, and obtaining sample physical properties of the digital fabric image includes:

[0015] S201, performing mechanical drape simulation on the digital fabric to obtain a first drape strength;

[0016] S202, for each physical property parameter, reducing the physical property parameter to half of the original value, and performing mechanical drape simulation on the digital fabric again to obtain a second drape strength;

[0017] S203, replacing the digital fabric, and using the physical property parameters to perform mechanical drape simulation on the replaced digital fabric to obtain a third drape strength;

[0018] S204, determining the significance of the influence of the physical property parameter on the mechanical drape of the digital fabric according to the first drape strength, the second drape strength and the third drape strength;

[0019] S205, selecting physical property parameters with higher impact significance, for the selected physical property parameters, determining the sampling density of the corresponding physical property parameters based on the impact significance, sampling the physical property parameters of the digital fabric based on the sampling density, and using the sampled physical property parameters as sample physical properties of the digital fabric image; wherein the sampling density is positively correlated with the impact significance.

[0020] Optionally, in S204, determining the significance of the influence of the physical property parameter on the mechanical drape of the digital fabric according to the first drape strength, the second drape strength, and the third drape strength includes:

[0021] S241, performing a difference calculation on the second draping strength and the first draping strength corresponding to each physical property parameter to obtain a property difference of each physical property parameter,

[0022] S242, performing a difference calculation on the third drape strength and the first drape strength corresponding to each physical property parameter to obtain a fabric difference of each physical property parameter,

[0023] S243, determining the influence significance of each physical property parameter based on the property difference and the fabric difference; wherein the property difference and the influence significance are positively correlated, and the fabric difference and the influence significance are negatively correlated.

[0024] Optionally, the method further includes:

[0025] Performing numerical adjustments on the sampled samples of the digital fabric image to obtain derivative samples;

[0026] The sampling samples and the derived samples are used as sample physical properties of the digital fabric image.

[0027] Optionally, in S300, determining the image region of the digital fabric image, and assigning the area, graphic features, and inflection points of the image region as image attributes to the digital fabric image includes:

[0028] S310, adjusting the size of the digital fabric image to a target size, determining an outer contour of the adjusted digital fabric image, and generating a vector graph of an image area within the outer contour;

[0029] S320, taking the center of the image area as the coordinate origin, establishing a plane coordinate system with the horizontal direction as the X axis and the vertical direction as the Y axis;

[0030] S330, determining the coordinates of each curve point on the outer contour line in the plane coordinate system, taking the curve point with the smallest X-axis coordinate in the first quadrant as the starting point, and numbering each curve point in a clockwise direction along the outer contour line in the vector diagram;

[0031] S340, calculating and summarizing the line segment slopes between two adjacent curve points, and generating a curvature data table of each curve point based on the line segment slopes;

[0032] S350, generating a curvature change graph according to the curvature data table, wherein the X-axis in the curvature change graph is the length of the outer contour, and the Y-axis is the curvature of the curvature point;

[0033] S360, determining the graphic features and inflection points of the curvature change graph: wherein the graphic features include the number of wrinkles, the depth of convex wrinkles, the depth of concave wrinkles, the width of convex wrinkles, and the width of concave wrinkles;

[0034] S370: Assign the area of ​​the image region, the graphic features and the inflection points of the curvature change graph as image attributes to the digital fabric image.

[0035] Optionally, in S400, constructing a training sample based on the image attributes of the digital fabric image and the sample physical attributes, inputting the training sample into a neural network for training, and obtaining a physical attribute recognition model for the digital fabric includes:

[0036] S410, dividing the physical properties of each sample into a plurality of levels arranged in descending order according to the order of influence significance from large to small, and determining the digital fabric image corresponding to the physical properties of the sample in each level,

[0037] S420, for each gear, obtaining sample physical properties and image properties of the digital fabric image in the gear as training samples,

[0038] S430, inputting the training samples of each gear into the neural network for training in sequence, and establishing the mapping relationship between the physical properties of each sample and the corresponding image properties in sequence, thereby obtaining a physical property recognition model of the digital fabric.

[0039] On the other hand, an embodiment of the present invention provides a digital fabric physical property recognition system based on pattern recognition technology, comprising:

[0040] at least one processor;

[0041] at least one memory for storing at least one program;

[0042] When the at least one program is executed by the at least one processor, the at least one processor implements the above method.

[0043] The beneficial effects of the present invention are as follows: the present invention discloses a method and system for identifying the physical properties of digital fabrics based on graphic recognition technology. The present invention constructs a digital fabric image library, determines the significance of the influence of each physical property parameter on the appearance of the fabric, and samples each physical property parameter accordingly based on the significance of the influence to obtain the sample physical properties of the digital fabric image; the accurate identification of the physical properties of the fabric can be achieved, and the efficiency and accuracy of fabric design can be improved. By determining the image area of ​​the digital fabric image, the area, graphic features and inflection points of the image area are assigned to the digital fabric image as image attributes; a training sample is constructed based on the image attributes and sample physical attributes of the digital fabric image, and the training sample is input into a neural network for training to obtain a physical property recognition model for the digital fabric; by training the neural network, the association between the fabric image attributes and the physical attributes is automatically established, which greatly improves the recognition efficiency. The present invention can accurately obtain the physical property parameters of the fabric without the need for traditional measuring tools. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0045] Figure 1 It is a flow chart of a method for identifying physical properties of digital fabrics based on pattern recognition technology according to an embodiment of the present invention;

[0046] Figure 2 is a top view of the fabric in an embodiment of the present invention;

[0047] Figure 3 This is a diagram showing the effect of different strengths on draping strength in an embodiment of the present invention;

[0048] Figure 4 : is a diagram showing the effect of different bending strengths on draping strength in an embodiment of the present invention;

[0049] Figure 5 is a bar graph of the bending strength value distribution in an embodiment of the present invention;

[0050] Figure 6 is a distribution line diagram of bending strength values ​​in an embodiment of the present invention;

[0051] Figure 7 is a histogram of intensity value distribution in an embodiment of the present invention;

[0052] Figure 8 is a distribution line diagram of intensity values ​​in an embodiment of the present invention;

[0053] Fig. 9 is a schematic diagram of a plane coordinate system in an embodiment of the present invention;

[0054] Fig.10 is a scatter plot of the slope of a line segment between two adjacent curve points in an embodiment of the present invention;

[0055] Fig.11 yes Fig.10 Schematic diagram of the graphic features in ;

[0056] Fig.12 It is a structural schematic diagram of a digital fabric physical property recognition system based on pattern recognition technology according to an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The following will be combined with the embodiments and drawings to clearly and completely describe the concept, specific structure and technical effects disclosed in the present invention, so as to fully understand the purpose, scheme and effect disclosed in the present invention. It should be noted that the embodiments and features in the embodiments in this application can be combined with each other without conflict.

[0058] The present invention proposes a digital fabric physical property recognition method and system based on graphic recognition technology. The method is based on artificial intelligence graphic recognition technology and can accurately obtain the physical property parameters of the fabric without the need for traditional measuring tools.

[0059] refer to Figure 1 ,like Figure 1 A method for identifying physical properties of digital fabrics based on pattern recognition technology provided by an embodiment of the present invention is shown, and the method comprises the following steps:

[0060] S100, constructing a digital fabric image library, wherein the digital fabric image library includes a plurality of digital fabric images, each of which is configured with corresponding physical property parameters;

[0061] S200, determining the significance of the influence of each physical property parameter on the appearance of the fabric, and sampling each physical property parameter accordingly based on the significance of the influence to obtain sample physical properties of the digital fabric image;

[0062] S300, determining an image region of the digital fabric image, and assigning the area, graphic features and inflection points of the image region as image attributes to the digital fabric image;

[0063] S400, constructing a training sample based on the image attributes and sample physical attributes of the digital fabric image, inputting the training sample into a neural network for training, and obtaining a physical attribute recognition model of the digital fabric; wherein the input data of the training sample is the image attribute of the digital fabric image, and the output data of the training sample is the sample physical attribute of the digital fabric image;

[0064] S500, collecting image data of real fabrics, preprocessing the collected image data, and inputting the data into a physical property recognition model to obtain predicted physical properties of the real fabrics.

[0065] Specifically, a database containing digital fabric images is constructed. The size of each fabric graphic is a circle with a radius of 13 cm and is placed at the center of a circular table with a radius of 8 cm and a height of 30 cm.

[0066] By setting a virtual lens at 20 cm from the top, the top view of the fabric is obtained. Figure 2 Each digital fabric image is equipped with corresponding physical property parameters, including warp, weft, diagonal strength, bending strength, deformation rate and deformation strength.

[0067] As shown in the table below;

[0068]

[0069]

[0070] In some embodiments, in S200, determining the significance of the impact of each physical property parameter on the fabric appearance, sampling each physical property parameter accordingly based on the significance of the impact, and obtaining the sample physical properties of the digital fabric image includes:

[0071] S201, performing mechanical drape simulation on the digital fabric to obtain a first drape strength;

[0072] S202, for each physical property parameter, reducing the physical property parameter to half of the original value, and performing mechanical drape simulation on the digital fabric again to obtain a second drape strength;

[0073] S203, replacing the digital fabric, and using the physical property parameters to perform mechanical drape simulation on the replaced digital fabric to obtain a third drape strength;

[0074] S204, determining the significance of the influence of the physical property parameter on the mechanical drape of the digital fabric according to the first drape strength, the second drape strength and the third drape strength;

[0075] S205, selecting physical property parameters with higher impact significance, for the selected physical property parameters, determining the sampling density of the corresponding physical property parameters based on the impact significance, sampling the physical property parameters of the digital fabric based on the sampling density, and using the sampled physical property parameters as sample physical properties of the digital fabric image; wherein the sampling density is positively correlated with the impact significance.

[0076] In this embodiment, after determining the significance of the influence of each physical property parameter on the mechanical drape of the digital fabric, the physical properties with higher significance can be determined and targeted sampling can be performed.

[0077] In some embodiments, in S204, determining the significance of the influence of the physical property parameter on the mechanical drape of the digital fabric according to the first drape strength, the second drape strength, and the third drape strength includes:

[0078] S241, performing a difference calculation on the second draping strength and the first draping strength corresponding to each physical property parameter to obtain a property difference of each physical property parameter,

[0079] S242, performing a difference calculation on the third drape strength and the first drape strength corresponding to each physical property parameter to obtain a fabric difference of each physical property parameter,

[0080] S243, determining the influence significance of each physical property parameter based on the property difference and the fabric difference; wherein the property difference and the influence significance are positively correlated, and the fabric difference and the influence significance are negatively correlated.

[0081] Specifically, in order to accurately quantify the impact of fabric physical property parameters on appearance, we conducted a series of experimental evaluations. Through these experiments, we were able to determine the specific impact of each physical property parameter on the appearance of the fabric, thereby providing the necessary data volume basis for building a digital fabric database. The experimental steps are as follows:

[0082] (1) Initial experiment: First, the mechanical drape of digital denim fabric was simulated and its drape characteristics were recorded.

[0083] (2) Adjustment of physical property parameters: Then reduce a certain type of physical property parameter of the denim fabric to twice the original value or reduce it to half of the original value, perform mechanical drape simulation again, and observe the change in drape.

[0084] (3) Material replacement: Then replace it with poplin fabric, adjust its physical property parameters to match the original strength of denim, and perform mechanical drape simulation.

[0085] Result comparison: Finally, by comparing the three groups of simulation results, the specific influence of a certain type of physical property parameters on the mechanical drape of digital fabrics is analyzed.

[0086] The specific operation process of experimenting with various physical property parameters is as follows:

[0087] 1. Strength: It includes four physical property parameters: weft yarn strength, warp yarn strength, diagonal tension (right), and diagonal tension (left). Generally, the four physical property parameters will increase simultaneously to indicate the stiff and inelastic physical properties of the fabric.

[0088] Conclusion: The experimental results show that for the same fabric, an increase in strength will lead to a corresponding increase in drape strength. However, due to the differences in their physical property parameters, different fabrics will show different drape strength even if they have the same strength. This shows that mechanical drape strength is not determined by strength alone, but is the result of the combined effect of multiple physical property parameters. Among these physical property parameters, strength has a more significant effect on drape strength.

[0089] 2. Bending strength: including bending strength-weft yarn, bending strength-warp yarn, bending strength-diagonal (right), bending strength-diagonal (left), which shows the stiffness of the fabric and the ease of bending.

[0090] Conclusion: The experimental results show that for the same fabric, the reduction of bending strength will lead to a very significant increase in drape strength. Different fabrics with the same bending strength will show differences in drape strength, but the differences are not very obvious. This shows that among these physical property parameters, bending strength has the most significant effect on drape strength.

[0091] 3. Deformation rate: including deformation rate-weft yarn, deformation rate-warp yarn, deformation rate-diagonal (right), deformation rate-diagonal (left), which shows the ratio of elongation or shortening of warp yarn when the fabric is stretched or compressed in the longitudinal direction.

[0092] Conclusion: The experimental results show that for the same fabric, doubling the deformation rate will not lead to a very significant increase in drape strength. Different fabrics with the same deformation rate will have very different drape strengths. This shows that among these physical property parameters, the deformation rate has no significant effect on drape strength.

[0093] 4. Deformation strength: including deformation strength-weft yarn, deformation strength-warp yarn, deformation strength-diagonal (right), deformation strength-diagonal (left), which are physical property parameters that show how easy the fabric is to bend.

[0094] Conclusion: The experimental results show that for the same fabric, doubling the deformation strength will lead to an increase in drape strength. Different fabrics with the same deformation strength will have very different drape strengths. This shows that among these physical property parameters, the deformation strength has a more significant effect on drape strength.

[0095] refer to Figure 3 and Figure 4 According to the influence of fabric physical properties on appearance, the influence is divided into three levels: most significant, relatively significant, and not significant. As shown in the following table:

[0096]

[0097] When collecting images, we will implement intensive gap settings for the most significant physical property parameters (bending strength) to ensure data coverage. On the contrary, for physical property parameters with more significant strength, we will adopt a looser gap setting to reduce unnecessary collection work. Although the influence of deformation strength is considered significant, its values ​​tend to be consistent for most conventional fabrics. For particularly stiff materials, such as stiff lining and pearl leather, the value of this physical property parameter may be close to 100. Therefore, we consider the setting of this physical property parameter to be standard and do not use it as an independent variable in the gap setting and image collection. Similarly, for physical property parameters with insignificant influence (deformation rate), we will not include it as an independent variable in the gap setting and image collection. This approach aims to optimize the collection process, ensure the effective use of resources, and improve the efficiency and quality of data collection. Through permutations and combinations, it was determined that 1,188 digital fabric data sets were needed to cover all attribute classifications.

[0098] The data collection of conventional digital fabric physical data is shown in the following table:

[0099]

[0100]

[0101] In some embodiments, the method further comprises:

[0102] Performing numerical adjustments on the sampled samples of the digital fabric image to obtain derivative samples;

[0103] The sampling samples and the derived samples are used as sample physical properties of the digital fabric image.

[0104] Specifically, 63 commonly used digital fabrics were collected and multiple jpg format images were obtained:

[0105] 1. Bending strength is obtained for sampling samples: When analyzing the bending strength of fabrics, we considered the strength in three directions: weft, warp, and diagonal. Research data shows that the bending strength value of the diagonal is usually between the bending strength of the weft and warp, which indicates that it can more comprehensively reflect the overall bending performance of the fabric. Therefore, we choose the bending strength of the diagonal as the main analysis indicator to ensure the accuracy and representativeness of the evaluation results. When analyzing the numerical distribution of bending strength, we use Figure 5 and Figure 6 It is observed that the bending strength of fabrics is mainly concentrated in the range of 41-55, and fabrics in this range account for the highest proportion and have a small range of numerical variation. It is followed closely by the range of 55-69, which accounts for the second largest proportion. There is also a certain proportion of fabrics in the range of 13-27, while the range of 27-41 is relatively small. As for fabrics with a bending strength of 97-100, due to their extremely stiff characteristics, they are usually used for specific auxiliary materials, so they are not considered in this analysis. Accordingly, we exclude this group of data from the analysis. The number of fabrics with bending strength values ​​in the range of 30-43 is small, and the numerical variations between fabrics are large, so the adjustment range of the numerical values ​​is large.

[0106] In order to enrich the sample samples in the database, we will adjust the bending strength values ​​of the existing 63 commonly used digital fabrics. In this way, we aim to generate a more diverse dataset and the corresponding digital fabric images. The specific steps and methods for adjusting the bending strength will be guided by the following Table 4. Such a strategy can not only increase the diversity of the database, but also provide a wider reference for the digital analysis of fabrics.

[0107] area Adjust the value The number of derivative samples added for each sample 13-22 ±2，±4 4 30-43 ±3，±6 4 48-69 ±1，±2 4

[0108] Since bending strength is the most significant physical property parameter, each inherent sampling sample will derive 4 samples. After the above operation, 63 regular sampling samples will derive 4 derived samples, that is, 63*4=256 samples will be derived, and the database has a total of 315 samples. Bending strength (weft yarn) and bending strength (warp yarn) will also change synchronously with the same value.

[0109] 2. Strength is obtained by sampling samples: the same consideration angle as the bending strength, we choose the diagonal strength as the main analysis indicator. Figure 7 and Figure 8 As shown in the figure, fabrics with diagonal values ​​between 78 and 97 are stiff accessories, so they are not considered in this analysis. Samples with diagonal values ​​between 2 and 40 have a small change range, so the adjustment range is large.

[0110] area Adjust the value Each sample increases the number of database samples 0-40 ±1 2 44-65 ±2 2

[0111] Since bending strength is a physical property parameter with a more significant impact, each inherent sample will derive 2 samples. After the above operation, 315 regular sampling samples will derive 2 derived samples, that is, 315*2=630 samples will be derived, and the database has a total of 945 samples.

[0112] In some embodiments, in S300, determining the image region of the digital fabric image and assigning the area, graphic features and inflection points of the image region as image attributes to the digital fabric image includes:

[0113] S310, adjusting the size of the digital fabric image to a target size, determining an outer contour of the adjusted digital fabric image, and generating a vector graph of an image area within the outer contour;

[0114] S320, taking the center of the image area as the coordinate origin, establishing a plane coordinate system with the horizontal direction as the X axis and the vertical direction as the Y axis;

[0115] S330, determining the coordinates of each curve point on the outer contour line in the plane coordinate system, taking the curve point with the smallest X-axis coordinate in the first quadrant as the starting point, and numbering each curve point in a clockwise direction along the outer contour line in the vector diagram;

[0116] S340, calculating and summarizing the line segment slopes between two adjacent curve points, and generating a curvature data table of each curve point based on the line segment slopes;

[0117] S350, generating a curvature change graph according to the curvature data table, wherein the X-axis in the curvature change graph is the length of the outer contour, and the Y-axis is the curvature of the curvature point;

[0118] S360, determining the graphic features and inflection points of the curvature change graph: wherein the graphic features include the number of wrinkles, the depth of convex wrinkles, the depth of concave wrinkles, the width of convex wrinkles, and the width of concave wrinkles;

[0119] S370: Assign the area of ​​the image region, the graphic features and the inflection points of the curvature change graph as image attributes to the digital fabric image.

[0120] Specifically, the specific steps of this embodiment are as follows:

[0121] Define the length: Use CAD software to open the digital fabric image and enlarge / reduce the image according to the size of the grid on the image. The size of the grid in 3D construction is 10CM*10CM;

[0122] Generate vector diagram: The outer contour of the pattern is traced according to the edge of the digital fabric image to generate a vector diagram. Since the outer edge of the vector diagram is a curve, curve points will be automatically generated according to the degree of curvature of the curve. More curve points will be generated for the curve with a small curvature radius, and fewer curve points will be generated for the curve with a small curvature radius.

[0123] Get area: Calculate the total area of ​​the graphic by using the vector image obtained by CAD software, and assign the area as an attribute to the digital fabric image;

[0124] Establish coordinates: Draw a circle with a radius of 8cm at the center of the figure (the default center of the figure by CAD) (simulating the shape and area of ​​the table), and define the center of the figure as the origin of the coordinates, establish the horizontal direction as the X axis, and the vertical direction as the Y axis; Fig. 9 shown.

[0125] Define the coordinates of the curve points: In the established coordinate system, first assign coordinates to each point on the curve. Then, sort these points. The starting point of the sorting is the point with the smallest X-axis coordinate in the first quadrant, which is marked as point 1. Starting from this starting point, each point on the curve is numbered clockwise along the outer contour of the vector diagram. This sorting method ensures that the order of the points matches their physical positions on the graph, which facilitates further analysis and processing.

[0126] like Fig.10 As shown, the slope of the line segment between two adjacent curve points is then calculated and summarized to generate a curvature data table for each point.

[0127] Generate a curvature change graph based on the curvature data table: the X-axis is the length of the outer contour, and the Y-axis is the curvature of the point;

[0128] Define the graphical features and special points of the curvature variation graph:

[0129] Define feature points: Fig.11 As shown in Figure 1, an inflection point is a specific point where the slope of a function curve changes, specifically when the slope changes from positive to negative or from negative to positive. In mathematics and graphical analysis, inflection points can be divided into two types:

[0130] 1. Inflection point (increase-decrease): When the slope of a function changes from increasing to decreasing, this inflection point is usually located near the peak of the graph. This means that before this point, the slope of the function is positive (that is, the function value increases as the independent variable increases), and after this point, the slope becomes negative (that is, the function value decreases as the independent variable increases).

[0131] 2. Inflection point (decreasing-increasing): When the slope of the function changes from decreasing to increasing, this inflection point is usually located near the trough of the graph. Before this point, the slope of the function is negative (that is, the function value decreases as the independent variable increases), and after this point, the slope becomes positive (that is, the function value increases as the independent variable increases).

[0132] These inflection points are very important in graphical analysis because they mark significant changes in the function's behavior, such as from growth to decline or from decline to growth, which is crucial to understanding the overall trend and characteristics of the function.

[0133] Define graphic features:

[0134] Number of wrinkles: In the analysis of function graphs, the number of inflection points is closely related to the concave and convex characteristics of the graph. Specifically:

[0135] 1. Inflection point (decrease-increase): When the slope of the function graph changes from decreasing to increasing, this inflection point corresponds to the convex fold in the graph. The convex fold refers to the local maximum area formed by the upward bending of the function graph at this point.

[0136] 2. Inflection point (increase-decrease): When the slope of the function graph changes from increasing to decreasing, this inflection point corresponds to a concave fold in the graph. A concave fold refers to the local minimum area formed by the downward bending of the function graph at this point.

[0137] According to the symmetry of the function graph and the properties of local extrema, the number of convex folds is equal to the number of concave folds. This means that in a continuous and smooth function graph, the number of inflection points (decreased-increased) is equal to the number of convex folds, and the number of inflection points (increased-decreased) is equal to the number of concave folds.

[0138] Depth of convex wrinkles: When analyzing the depth of convex wrinkles in the function graph, we focus on each inflection point (increase-decrease), which marks the turning point where the function graph changes from an upward trend to a downward trend. In order to quantify the depth of each convex wrinkle, we select 10 consecutive coordinate points after each inflection point as the research object. Among these 10 points, the point farthest from the origin is defined as the peak point of the convex wrinkle. The depth of the convex wrinkle can be determined by calculating the straight-line distance between the peak point and the origin of the graph, and then subtracting the circular radius of 18 cm from it.

[0139] Depth of concave folds: When studying the depth of concave folds in the function graph, we take each inflection point (decreasing-increasing) and the 10 coordinate points after it as the analysis object. The inflection point (decreasing-increasing) marks the turning point of the graph from a downward trend to an upward trend. Among these 10 points, the point with the closest straight-line distance to the origin is identified as the trough point of the concave fold. The depth of the concave fold is calculated by measuring the straight-line distance between the trough point and the origin, and then subtracting the circular radius of 18 cm from this distance.

[0140] Width of raised wrinkles: The X-axis distance between the two inflection points (increase-decrease) is the width of the peak, i.e. the width of the raised wrinkles;

[0141] Concave fold width: The X-axis distance between the two inflection points (decrease-increase) is the width of the trough, that is, the width of the concave fold.

[0142] In some embodiments, in S400, constructing a training sample based on the image attributes of the digital fabric image and the sample physical attributes, inputting the training sample into a neural network for training, and obtaining a physical attribute recognition model for the digital fabric includes:

[0143] S410, dividing the physical properties of each sample into a plurality of levels arranged in descending order according to the order of influence significance from large to small, and determining the digital fabric image corresponding to the physical properties of the sample in each level,

[0144] S420, for each gear, obtaining sample physical properties and image properties of the digital fabric image in the gear as training samples,

[0145] S430, inputting the training samples of each gear into the neural network in sequence for training, and establishing the mapping relationship between the physical properties of each sample and the corresponding image properties in sequence, thereby obtaining a physical property recognition model of the digital fabric.

[0146] This step is the AI ​​learning stage, which uses a graded digital fabric library to train AI to identify the patterns of graphic changes and their corresponding changes in physical property values.

[0147] Specifically, since the digital fabrics in the digital fabric library are divided into grades, which means that they do not change continuously, artificial intelligence learning is required to derive the pattern change rules between the two grades and the corresponding physical value changes;

[0148] Data processing: Perform the graphic processing and data definition data of the previous step on the 945 data; and assign the physical attribute value corresponding to each image as an attribute to the image (bending strength, strength);

[0149] Arrange the research objects according to the gears, first arranging the factors with the most significant shape influence (bending strength), then the factors with relatively significant shape influence (strength and deformation strength), and finally the factors with insignificant shape influence (deformation rate), and carry out artificial intelligence stage learning in this order.

[0150] First level learning:

[0151]

[0152]

[0153] Second level learning:

[0154]

[0155] The third level of learning: Use the same method as above to learn the relationship between the highest convex height of the fold and the bending strength and strength;

[0156] Fourth level of learning: Use the same method as above to learn the relationship between the minimum depression height of the fold and the bending strength and strength.

[0157] In this embodiment, a plurality of digital fabric images are randomly selected from the digital fabric images in each gear to construct a training set. First, the area feature of the image is identified, followed by the wave number feature, then the wave peak value, and finally the wave trough value.

[0158] Virtual verification and training, including pattern recognition training in the database, pattern recognition training outside the database, and real fabric verification training;

[0159] Image recognition training in the database: During the training of the convolutional neural network (CNN), we randomly selected 100 images from the existing 945 images in the gallery to build a training set. The goal of the training is to enable the model to recognize and accurately match images that are exactly the same as the query image. During the training of the convolutional neural network (CNN), we set a specific recognition order for the model, with the priority being: first recognize the area feature of the image, then the wave number feature, then the peak value, and finally the trough value, including:

[0160] 1. Area: Locate 10 images in the database that have the same or similar areas as the known image;

[0161] 2. Wave number: Match 5 pictures with the same or similar wave numbers as the known figures among these 10 pictures;

[0162] 3. Maximum convex wrinkles (wave crests): Find 3 pictures with the same or similar maximum convex wrinkles (wave crests) as the known figures among the 5 pictures;

[0163] 4. Minimum concave wrinkles (troughs): Find the two pictures with the same or similar maximum convex wrinkles (peaks) as the known figures among the three pictures;

[0164] 5. Manual selection: manually select one picture from the two pictures. This action serves as data guidance for computer self-learning.

[0165] According to the training results, it can be confirmed that the convolutional neural network (CNN) can accurately perform image recognition according to the established requirements. This shows that the CNN model has successfully learned the feature representation of the image and can accurately recognize the target image in practical applications.

[0166] Graphics recognition training outside the database: In order to train the convolutional neural network (CNN) for pattern recognition, we first artificially set a series of bending strength values ​​for fabrics and generated 100 new digital fabric images based on these values. These images have never appeared in the database. Then, we followed a specific recognition step, that is, the pattern recognition training process in the database, to train the CNN model so that it can recognize these new fabric images and give the corresponding physical properties. In this way, the results of the neural network learning phase can be verified, and if the results are not ideal, it can be learned again;

[0167] Real fabric verification training: During the training process, we collected 50 samples of real fabric images, whose physical property parameters have been obtained by professional measuring instruments. We used the image processing method previously used to process digital fabrics to analyze these real fabric images and extracted the area, wave number, maximum convex wrinkle value and minimum concave wrinkle value of the images. Then, we determined the physical property values ​​of the fabrics according to the recognition order in the image recognition training outside the database, and found the images closest to these real fabric images from the virtual fabric image library. Finally, we compared the differences between the physical property values ​​obtained by the image processing method and the values ​​measured by professional measuring instruments, and evaluated the similarity between the real fabric images and the virtual fabric images located by the system to verify the effectiveness of the training.

[0168] Use of this method:

[0169] In order to simulate real fabrics and try on samples in a virtual environment, we need to obtain the physical properties of the fabrics. To do this, we collect image data of the fabrics according to the specified shooting requirements and perform subsequent processing on the collected images. The processed images and their data will be used to extract the physical properties of the fabrics, including their bending strength values. These physical property values ​​are then applied to the fabric simulation physical property module in the virtual simulation software to build a digital version of the fabric and perform a simulated try-on to evaluate its performance in the virtual sample. This process ensures that the physical properties of the fabrics can be accurately reflected in the virtual design.

[0170] Initial completion and continuous optimization of the model:

[0171] After the initial model is built, the accuracy of the model and the authenticity of the physical properties of the simulation effect can be improved by continuously adding new digital fabric data, such as different shooting angles and size specifications, such as adding shooting angles, adding fabric specifications or changing the display stand.

[0172] Through the above steps, the present invention can not only realize the rapid and accurate acquisition of the physical properties of digital fabrics, but also provide an efficient and low-cost solution for the field of fabric design and manufacturing.

[0173] Compared with the prior art, the embodiments provided by the present invention have the following improvements:

[0174] A method is proposed to obtain the physical property values ​​required for virtual simulation of fabrics by comparing the fabric draping shape pictures with the digital fabric virtual simulation picture database.

[0175] A classification of the influence of various physical property parameters on the appearance of fabrics is proposed during the virtual simulation stage.

[0176] A method for extracting features from digital fabric images is proposed: area, curvature change, fold depth, and fold width.

[0177] It proposes artificial intelligence technology combined with image extraction features to learn the changing rules of the impact of changes in various physical attribute parameters on appearance.

[0178] refer to Fig.12 The embodiment of the present invention further provides a digital fabric physical property recognition system based on graphic recognition technology, comprising:

[0179] at least one processor;

[0180] at least one memory for storing at least one program;

[0181] When the at least one program is executed by the at least one processor, the at least one processor implements the above method.

[0182] The contents of the above method embodiments are all applicable to this embodiment. The functions specifically implemented by this embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments, which will not be repeated here.

[0183] Although the description of the present disclosure has been quite detailed and specifically describes several described embodiments, it is not intended to be limited to any of these details or embodiments or any particular embodiment, but should be regarded as providing a broad possible interpretation of these claims by reference to the appended claims, taking into account the prior art, so as to effectively cover the intended scope of the present disclosure. In addition, the above description of the present disclosure is based on the embodiments foreseeable by the inventor, and its purpose is to provide a useful description, and those non-substantial changes to the present disclosure that have not yet been foreseen may still represent equivalent changes to the present disclosure.

Claims

1. A method for identifying physical properties of digital fabrics based on pattern recognition technology, characterized in that: The method comprises the following steps: S100, constructing a digital fabric image library, wherein the digital fabric image library includes a plurality of digital fabric images, each of which is configured with corresponding physical property parameters; S200, determining the significance of the influence of each physical property parameter on the appearance of the fabric, and sampling each physical property parameter accordingly based on the significance of the influence to obtain sample physical properties of the digital fabric image; S300, determining an image region of the digital fabric image, and assigning the area, graphic features and inflection points of the image region as image attributes to the digital fabric image; S400, constructing a training sample based on the image attributes and sample physical attributes of the digital fabric image, inputting the training sample into a neural network for training, and obtaining a physical attribute recognition model of the digital fabric; wherein the input data of the training sample is the image attribute of the digital fabric image, and the output data of the training sample is the sample physical attribute of the digital fabric image; S500, collecting image data of real fabrics, inputting the collected image data into a physical property recognition model, and obtaining predicted physical properties of the real fabrics.

2. According to claim 1, a method for identifying physical properties of digital fabrics based on pattern recognition technology is characterized in that: In S200, determining the significance of the influence of each physical property parameter on the appearance of the fabric, sampling each physical property parameter accordingly based on the significance of the influence, and obtaining the sample physical properties of the digital fabric image include: S201, performing mechanical drape simulation on the digital fabric to obtain a first drape strength; S202, for each physical property parameter, reducing the physical property parameter to half of the original value, and performing mechanical drape simulation on the digital fabric again to obtain a second drape strength; S203, replacing the digital fabric, and using the physical property parameters to perform mechanical drape simulation on the replaced digital fabric to obtain a third drape strength; S204, determining the significance of the influence of the physical property parameter on the mechanical drape of the digital fabric according to the first drape strength, the second drape strength and the third drape strength; S205, selecting physical property parameters with higher impact significance, for the selected physical property parameters, determining the sampling density of the corresponding physical property parameters based on the impact significance, sampling the physical property parameters of the digital fabric based on the sampling density, and using the sampled physical property parameters as sample physical properties of the digital fabric image; wherein the sampling density is positively correlated with the impact significance.

3. The method for identifying physical properties of digital fabrics based on pattern recognition technology according to claim 2, characterized in that: In S204, determining the significance of the influence of the physical property parameter on the mechanical drape of the digital fabric according to the first drape strength, the second drape strength and the third drape strength includes: S241, performing a difference calculation on the second draping strength and the first draping strength corresponding to each physical property parameter to obtain a property difference of each physical property parameter, S242, performing a difference calculation on the third drape strength and the first drape strength corresponding to each physical property parameter to obtain a fabric difference of each physical property parameter, S243, determining the influence significance of each physical property parameter based on the property difference and the fabric difference; wherein the property difference and the influence significance are positively correlated, and the fabric difference and the influence significance are negatively correlated.

4. The method for identifying physical properties of digital fabrics based on pattern recognition technology according to claim 3, characterized in that: The method further comprises: Performing numerical adjustments on the sampled samples of the digital fabric image to obtain derivative samples; The sampling samples and the derived samples are used as sample physical properties of the digital fabric image.

5. The method for identifying physical properties of digital fabrics based on pattern recognition technology according to claim 1, characterized in that: In S300, determining the image region of the digital fabric image and assigning the area, graphic features and inflection points of the image region as image attributes to the digital fabric image includes: S310, adjusting the size of the digital fabric image to a target size, determining an outer contour of the adjusted digital fabric image, and generating a vector graph of an image area within the outer contour; S320, taking the center of the image area as the coordinate origin, establishing a plane coordinate system with the horizontal direction as the X-axis and the vertical direction as the Y-axis; S330, determining the coordinates of each curve point on the outer contour line in the plane coordinate system, taking the curve point with the smallest X-axis coordinate in the first quadrant as the starting point, and numbering each curve point in a clockwise direction along the outer contour line in the vector diagram; S340, calculating and summarizing the line segment slopes between two adjacent curve points, and generating a curvature data table of each curve point based on the line segment slopes; S350, generating a curvature change graph according to the curvature data table, wherein the X-axis in the curvature change graph is the length of the outer contour, and the Y-axis is the curvature of the curvature point; S360, determining the graphic features and inflection points of the curvature change graph: wherein the graphic features include the number of wrinkles, the depth of convex wrinkles, the depth of concave wrinkles, the width of convex wrinkles, and the width of concave wrinkles; S370: Assign the area of ​​the image region, the graphic features and the inflection points of the curvature change graph as image attributes to the digital fabric image.

6. The method for identifying physical properties of digital fabrics based on pattern recognition technology according to claim 1, characterized in that: In S400, the step of constructing a training sample based on the image attributes of the digital fabric image and the sample physical attributes, and inputting the training sample into a neural network for training to obtain a physical attribute recognition model for the digital fabric includes: S410, dividing the physical properties of each sample into a plurality of levels arranged in descending order according to the order of influence significance from large to small, and determining the digital fabric image corresponding to the physical properties of the sample in each level, S420, for each gear, obtaining sample physical properties and image properties of the digital fabric image in the gear as training samples, S430, inputting the training samples of each gear into the neural network for training in sequence, and establishing the mapping relationship between the physical properties of each sample and the corresponding image properties in sequence, thereby obtaining a physical property recognition model of the digital fabric.

7. A digital fabric physical property recognition system based on pattern recognition technology, characterized in that: include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method according to any one of claims 1 to 6.

Images