Fabric evaluation method and device, storage medium and computer equipment

Abstract

The invention discloses a fabric evaluation method and device, a storage medium and computer equipment, and relates to the technical field of fabric curvature simulation evaluation, and the method comprises the steps: building a multilayer neural network model; and processing the overhanging image by using the multi-layer neural network model to complete fabric evaluation. The method comprises the following steps: processing an obtained fabric overhanging image through a multi-layer neural network, filtering interference information through preprocessing, intercepting the inner and outer contours of the fabric, expanding the contours into waveforms, analyzing waveform parameters, feeding back the actual overhanging characteristics of the fabric, and further obtaining the bending hardness of the fabric. According to the invention, the multi-layer neural network model can process the overhanging images of various fabrics in a more refined manner, thereby achieving the precise control of the features of the fabrics, and facilitating the related workers to know the actual features of different types of fabrics.

Description

Technical Field

[0001] The present invention relates to the technical field of fabric bending degree simulation and evaluation, and particularly relates to a fabric evaluation method, device, storage medium, and computer device. Background Art

[0002] Bending hardness is an important criterion for measuring the quality of fabrics, but it is extremely difficult to be accurately measured and simulated. Performance and accuracy are two important indicators for measuring physical fabric simulation tests; although there have been great improvements in fabric simulation in recent years, there has been little progress in the accuracy of fabric simulation. In order to be able to restore the real state of fabrics as much as possible, it is necessary to place real clothing fabrics more accurately into simulation software for simulation. Among the many factors affecting simulation accuracy, the planar hardness and bending hardness of fabrics are two of the most critical factors. Planar hardness is usually only important for elastic fabrics, and bending hardness determines the softness of fabrics and the details of wrinkles. Therefore, bending hardness is almost an important indicator for measuring all fabrics. However, the bending hardness of fabrics has characteristics such as non-linearity, anisotropy, and diversity. Therefore, how to accurately restore the bending hardness of fabrics in simulation software has become a rather difficult problem. The methods for measuring the bending hardness of fabrics include the cantilever method, the heart method, and the draping method. Among these, the cantilever method is the most common and can most intuitively reflect the fabric characteristics. The working principle of the cantilever method is to use a cantilever in the air to test the bending degree of a fabric strip under its own weight. In the field of graphics, the cantilever method has also become a method for obtaining the standard of bending hardness parameters. However, the prior art cannot process the draping image carefully and correctly when processing the draping image, resulting in the accuracy of the subsequent determination of the bending hardness of the fabric being affected. Therefore, how to accurately process the draping image is an urgent problem to be solved for accurately obtaining the bending hardness of fabrics nowadays.

[0003] Chinese Patent Document CN114925600A discloses "A learning-based method for measuring the bending hardness of draped fabrics". It obtains the relationship between the real fabric data and the non-linear bending modulus and the anisotropic bending modulus, and constructs a parameter data set; normalizes the parameters in the parameter data set to obtain a processed parameter data set; constructs a VAE subspace model using the processed parameter data set; obtains the initial state of each parameter vector in the VAE subspace model to generate a simulated data set; generates multi-view depth maps through the simulated data set; uses the multi-view depth maps to obtain a learned deep neural network through the learning of the deep neural network; and uses the learned deep neural network to obtain the bending hardness of the real fabric to be measured. The "obtaining the relationship between the real fabric data and the non-linear bending modulus and the anisotropic bending modulus" includes: preparing a sample of the real fabric and obtaining an image of the sample; obtaining a set of curve sample points on the image according to the image, and obtaining the moment of any curve sample point in the set of curve sample points; and obtaining the non-linear bending modulus and the anisotropic bending modulus of the corresponding real fabric according to the moment of any curve sample point. However, in the process of simulating the drape of the fabric curvature, this patent cannot process the drape image more precisely. Summary of the Invention

[0004] The present invention mainly solves the technical problem of poor processing of drape images in the original simulation test of fabric curvature, and provides a fabric evaluation method, device, storage medium, and computer device.

[0005] The above technical problems of the present invention are mainly solved by the following technical solutions: The present invention includes S1: Establish a multi-layer neural network model to obtain the fabric drape image and perform parameterization processing; S2: Extract the image contour of the fabric drape image and unfold it to complete the fabric evaluation.

[0006] As the types of raw materials for making clothing are becoming increasingly rich, the characteristics of clothing fabrics are also becoming increasingly complex, which makes it difficult for workers to grasp the actual characteristics of clothing. Using the old-fashioned manual testing method to evaluate fabric characteristics requires a large amount of resource costs. In order to reduce the testing costs, most existing methods simulate the actual fabric by establishing a simulation model for fabric bending hardness testing. However, in the process of fabric bending hardness testing using the existing simulation models, the processing of fabric drape images is usually ignored. The accuracy of image processing can directly affect the subsequent simulation model's judgment of fabric bending hardness and actual fabric characteristics. This patent accurately processes the drape image obtained through a multi-layer neural network model, analyzes and obtains the fabric bending hardness characteristics from the image data through multiple processes, so as to accurately evaluate and restore the actual characteristics of the fabric.

[0007] Preferably, in the step S1, a number of fabric control parameters are preset in advance. After obtaining the parameters, the input layer, output layer and hidden layer of the neural network are constructed. The number of layers, nodes and activation function types of the input layer, output layer and hidden layer are set respectively. The actual fabric photographed image is used as the input vector of the multi-layer neural network, and the fabric bending hardness is used as the output vector of the multi-layer neural network to realize the mapping from the fabric image to the fabric draping parameters. Under the combined action of the multi-layer neural network, the real-time photographed fabric test image is imported and converted into the fabric curvature feature, realizing the process of automatically processing the image into feature parameters.

[0008] Preferably, in the step S2, after the multi-layer neural network model is established, the draping image parameterization process is carried out according to the fabric test image and the model structure obtained by the multi-layer neural network model. The image area after parameterization of the preprocessing of the draping image is calculated, the image contour of the draping image is extracted and unfolded, and peak point sampling is carried out on the unfolded image to obtain the draping parameters. After the above steps of processing, relevant personnel can know the bending hardness characteristics of the processed fabric by means of the draping parameter part in the image, reducing the labor cost and time cost required for manual detection.

[0009] Preferably, in the step S2, the draping image parameterization process includes at least one of image preprocessing, image area calculation, image contour extraction, image contour unfolding, peak point extraction and draping parameter extraction to obtain the draping index of the virtual fabric.

[0010] Preferably, in the step S2, the image preprocessing step includes converting the original fabric draping image into a grayscale image and taking the inverse color, and the inverse color image is segmented after filtering. Image preprocessing can remove redundant interference feature information on the image, facilitating the multi-layer neural network model to more intuitively process and obtain the draping image processing result.

[0011] Preferably, in the step S2, the image is converted into a binary image and the image area is calculated. When calculating the area, the pixel value of the point in the draping image area is represented as 1, and the value of the background is represented as 0. The calculation method of the image area is: A = ∑ (x，y)∈R f(x, y), where R is the set of pixels in the draping image area, x and y are the two-dimensional coordinates of the draping image, and the number of statistical representations is the area of the draping image. Distinguishing the background part and the fabric draping projection part in the image according to the pixel value representation is convenient for subsequent segmentation and extraction of the pixel features of the fabric part in the whole image, further eliminating interference information and improving the accuracy of the processing result of the multi-layer neural network model.

[0012] Preferably, in the step S2, after the above processing of the drape image, the following steps are further included: extracting and collecting the drape image contour, collecting the drape image contour, and extracting the internal edge formed by the internal contour holes of the drape image while extracting the drape external contour; unfolding the drape image contour, constructing a graphic circle with the center of the fabric drape image as the center, determining the radius and angle of the graphic circle, and unfolding and extending the image contour according to the radius and angle of the circle. The shapes of the fabrics to be detected are not all in the case of solid interiors and regular edges. There are cases where the external shapes are complex and there are holes inside the fabrics. Therefore, when collecting the image contour, it is necessary to restore the internal and external morphological characteristics of the fabric synchronously to more accurately test the bending hardness characteristics of the fabric.

[0013] Preferably, in the step S2, after the above processing of the drape image, its contour is extracted and collected. The image contour is collected, and the internal edge formed by the internal contour holes of the image is extracted while extracting the external contour; the image contour is unfolded, a graphic circle is constructed with the center of the fabric drape image as the center, the radius and angle of the graphic circle are determined, and the image contour is unfolded and extended according to the radius and angle of the circle. After obtaining the fabric characteristics, in order to further reflect the bending characteristics of the fabric contour, a waveform is obtained by unfolding at an angle after determining the center of the circle, and the bending hardness of fabrics of various shapes can be feedback from the waveform through subsequent processing steps.

[0014] Preferably, in the step S2, the peak point sampling and drape parameter extraction include calculating the static drape coefficient, and the calculation method of the static drape coefficient is: where A_1 is the projected area of the fabric, A_2 is the area of the fabric tray, and A_3 is the area of the fabric.

[0015] Preferably, in the step S2, the peak point sampling step further includes: wave number collection. After the drape projection contour of the fabric is unfolded, there are peaks or valleys, and the number of peaks and valleys is counted to obtain the wave number; confirmation of the peak angle, the peak angle refers to the angle between two adjacent peaks of the fabric projection contour; determining the peak amplitude according to the obtained peak situation, and the peak amplitude mainly determines the distance from the peak point in the fabric drape projection to the edge of the fabric tray. Analyze the waveform of the unfolded fabric contour, and assist relevant personnel in grasping the bending hardness of the fabric to be tested from aspects such as the waveform related parameters of the static drape coefficient.

[0016] Preferably, a fabric evaluation device includes an image acquisition module for acquiring real-time images of fabric bending hardness tests; a neural network module for processing fabric drape images and obtaining drape image processing results; and an evaluation module for displaying fabric bending degree simulation test results, evaluating the fabric bending degree, and feedbacking the results.

[0017] Preferably, a storage medium includes a draped image processing program, which when executed, causes the computer to execute according to a predetermined draped image processing method.

[0018] A computer device includes a memory and a processor. The memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory. It is characterized in that when the computer program runs, the processor executes the fabric evaluation method according to any one of claims 1-8.

[0019] The beneficial effects of the present invention are as follows: The draped fabric images obtained are processed by a multi-layer neural network. After preprocessing to filter out interference information, the internal and external contours of the fabric are intercepted and unfolded into waveforms. The waveform parameters are analyzed to feedback the actual draping characteristics of the fabric, and thus the bending hardness of the fabric is obtained. Through the multi-layer neural network model of the present invention, the draped images of various fabrics can be processed more precisely, thereby realizing the accurate control of fabric characteristics and facilitating relevant staff to understand the actual characteristics of different types of fabrics. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a flowchart of a fabric evaluation of the present invention.

[0021] Figure 2 is a flowchart of peak point sampling of the present invention.

[0022] Figure 3 is a flowchart of image segmentation of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] The technical solutions of the present invention will be further specifically described below through embodiments in conjunction with the drawings.

[0024] Example 1: A fabric evaluation method, device, storage medium, and computer device in this example. In the current era of rapid development of information technology, digital information-related technologies are becoming increasingly mature. In the trend of the clothing industry's transformation by integrating digital information, the production process in the existing clothing industry shows a trend towards small batches, diversification, digitization, and timeliness. With the digital development of the clothing industry, clothing manufacturing has gradually entered the era of intelligent manufacturing in the industry, fully integrating the clothing industry into the clothing production process. However, when clothing enterprises carry out industrial intelligent manufacturing reforms, how to keep up with the trend of market changes and timely provide products that meet customer needs with low cost and high quality is the current dilemma they face. In the traditional clothing production process, before mass-producing clothing, it is necessary to conduct fabric property tests and sample clothing trials at least two or three times, and at most a dozen times. Multiple physical property tests of clothing fabrics can increase the probability of the finished clothing passing the review of designers or customers. The time consumed in this link accounts for a very high proportion in the product development cycle. If accurate fabric property evaluation cannot be achieved, it will seriously affect the production efficiency of subsequent finished clothing and greatly increase production costs. Therefore, in the process of intelligent manufacturing in the clothing industry, how to conduct fabric physical property inspection through a digital information model has become an extremely effective means. This method can test the relevant properties of various clothing fabrics through relevant models, thus transferring the physical property test of collecting various clothing fabrics in the real world to the digital model. Quickly testing the properties of various fabrics through the digital model can significantly shorten the clothing preparation and production cycle, thereby reducing production costs. The digital model can display the physical properties of the fabric in a three-dimensional state, and can give designers and pattern makers a reference before the real finished clothing is made, improving the efficiency from fabric control to sample clothing production to determining the finished clothing pattern, thereby significantly increasing the success rate of the finished clothing and greatly reducing the clothing production cycle time. However, there are still many problems in the existing process of detecting the physical properties of clothing fabrics through digital models, and the biggest problem is how to accurately identify the physical properties of different fabrics through digital models. The physical properties of virtual fabrics include the drape of virtual fabrics, the color and texture of virtual fabrics, the luster of virtual fabrics, etc. Among them, the drape of the fabric is the main attribute during the wearing process of clothing, which can intuitively show the difference between virtual clothing and real clothing. The drape of the fabric refers to the property that the fabric can produce a smooth and uniformly curved surface under the natural hanging state. When using a digital model to detect the drape of the fabric, how to accurately simulate and predict the drape of the fabric greatly affects the accuracy of subsequent finished clothing production. In the currently commonly used digital models, the drape of clothing is controlled by adjusting the physical property control parameters of virtual fabrics. However, the physical property control parameters of virtual fabrics have no unit, which requires relevant staff to rely on their experience and feeling to adjust the parameters to improve the accuracy of the digital model's simulation of the physical properties of fabrics, and it is also difficult to simulate an effect highly consistent with real fabrics.Therefore, in this embodiment, the detection accuracy of the physical properties of the fabric is improved by using a multi-layer neural network model. For example. Figure 1 As shown, it includes a multi-layer neural network model for establishing the refined image features of the fabric drape. The multi-layer neural network model in this embodiment includes an input layer, which is the access port of the entire multi-layer neural network, realizing the import of the drape image into this multi-layer neural network model, and taking the drape coefficient, peak amplitude, wave number, and peak angle of the drape image as the nodes of the input layer. The establishment of the input layer facilitates the subsequent determination of the relationship between the fabric drape parameters and the fabric bending hardness; the output layer is the external interface of the multi-layer neural network, realizing the export from the drape image to the fabric bending hardness characteristics, and the bending hardness of the fabric is the node of this layer; the hidden layer is the level where the main training work of the neural network model is carried out. The number of hidden layer nodes can be appropriately increased according to the training variables, which can significantly improve the training effect; after determining the layers of the neural network model, it is necessary to determine the selection of an activation function to assist the model in adjusting the error during the training process. The activation function generally uses a non-linear function with differentiable characteristics. The three layers of the neural network cooperate with each other and adjust the deviation of the training results with the help of the activation function, ultimately realizing the mapping from the fabric drape image parameters to the fabric bending hardness. In the existing process of testing the fabric bending hardness, most only pay attention to how to automatically simulate the process of detecting the fabric characteristics by the cantilever method through the model, but ignore the influence of the image processing accuracy of the drape in the testing process on the results. In this embodiment, a multi-layer neural network is used to simulate the test of the fabric bending hardness, and multi-segment processing is performed on various forms of drape images generated to exclude interference information and more accurately reflect the relationship between the parameters related to the drape image and the fabric.

[0025] Take the sag images of various fabrics measured by the cantilever method, and import the captured sag images into a multi-layer neural network model for parameterization. The parameterization steps include image preprocessing, calculating the image area, extracting and unfolding the image contour after area processing, and extracting peak points and calculating sag parameters after completing the image contour processing. When the fabric is tested by the cantilever method, the fabric will naturally hang down in a sagging shape, and the sagging degree and state of the fabric are the test indicators for this fabric. In this embodiment, the sag coefficient, wave number, wave peak amplitude, and wave peak angle of the fabric under the sagging state are used as the measurement indicators for the bending hardness of the fabric. Image preprocessing is to convert the R, G, and B components of the fabric sag image into the format of a grayscale image by the weighted average method for the true color image, multi-view depth map, and point cloud data image generated during the fabric cantilever test. Converting the original image format into the grayscale image format can reduce the image occupied space and improve the model running speed. The grayscale image after color inversion processing will generate interference information such as noise in the above processing steps, which will further lead to the deterioration and blurring of the image quality, and even the sagging features will disappear, ultimately causing the image to be unable to proceed with the subsequent processing steps. Therefore, after the color inversion processing of the grayscale image, the termination filtering method is used for filtering, and then the image after filtering the noise is segmented to extract the image part with fabric sagging features. The binaryzation method is used for image segmentation. Due to the differences in the number of peaks and processing processes in the grayscale image, there are also certain differences in the binaryzation processing methods. The commonly used binaryzation processing methods include the single-threshold method and the double-threshold method. Since the maximum between-class variance method of the single-threshold method has the advantage of a small misclassification probability, in this embodiment, the single-threshold method is used to obtain the optimal threshold, and the image is segmented by the between-class variance between the two. During segmentation, a threshold X is preset in advance, and the image is divided into two pixel groups with this threshold as the boundary, and they are respectively set to 0 and 1 for storage. After obtaining the total average gray value of the image, traverse the gray value in the entire image that is closest to the total average gray value as the candidate optimal gray threshold, and then calculate the second candidate optimal gray threshold by the optimal between-class variance method. Compare the above two gray values with the total average gray value, and the one with the smaller error is the threshold for image segmentation, and the image is cut with this threshold. Through the image segmentation step, the image part with fabric sagging features can be filtered out from the entire sag image as much as possible, excluding the interference of the noise of the redundant image part on the subsequent fabric sag parameter features, improving the processing accuracy of the multi-layer neural network model for the fabric sag image, and ultimately assisting the relevant staff to more accurately master the bending hardness characteristics of different fabrics.

[0026] After image segmentation, the sag image is a binary image. When calculating the area of this image, it is determined that the pixel value of the points in the sag image area is represented as 1, and the pixel value of the background is represented as 0. The calculation method of the image area is: A = ∑ (x，y)∈Rf(x, y), where R is the set of pixels in the draped image area, x and y are the two-dimensional coordinates of the draped image, and the number of statistical characterizations is the area of the draped image. During the process of obtaining the draping parameters by processing the draped image, most of the information of the image is often concentrated on the edge contour of the image. By means of the mutation points of the edge contour position signal, the position of the image contour can be obtained. Therefore, after determining the overall draped projection area of the fabric, it is necessary to further judge the draping characteristics of the fabric by detecting and extracting the image contour, so as to more fully perceive the bending hardness of the fabric. The extraction of the image contour requires obtaining the coordinate values of each point on the image for sampling. When encountering the singular points and mutation points in the image contour, the point is set as an image edge point, and the change of gray level is characterized by the change of the gray level distribution of the surrounding pixels, so as to complete the extraction of the image contour. After the extraction of the image contour is completed, the contour is unfolded in the form of a waveform to form a waveform. By studying the characteristics of this waveform, the bending hardness characteristics of the fabric are reflected. In this embodiment, when unfolding the image contour, first locate the center of the circle of the image contour, use the center point of the overall contour as the center of the circle, and establish a two-dimensional coordinate system with this center of the circle as the origin, and then determine the coordinates of each point on the contour, as well as the distance and angle from the center of the circle. After determining the above parameters, then preset the unfolding angle α, and use the angle as the abscissa and the distance as the ordinate to generate a waveform unfolding diagram of the image contour. Due to the irregularity of the fabric itself contour and the discreteness of digital technology, the initially unfolded waveform diagram will inevitably generate strange shapes. The waveform unfolding diagram of the initially generated image contour is smoothed by using the smooth function to improve the waveform characteristics, thereby improving the accuracy of the multi-layer neural network model for image processing.

[0027] After the image contour is unfolded, the corresponding function can be called to complete the extraction of peak points. Since the function for retrieving peak points may have multiple extreme points within a single period, in such cases, the function will judge multiple extreme values as peak points, which will cause the model to misjudge the peak points of the unfolded image contour, resulting in incorrect parameter judgment of the subsequent draped image, and further affecting the final judgment of the fabric bending hardness, and hindering the relevant staff's understanding and judgment of the fabric. Therefore, it is necessary to set the minimum height and minimum interval number of the wave peaks. By setting the minimum height of the wave peak value, the function can be prevented from misreading the trough points, and the same minimum height value of the peak will also be adjusted correspondingly with the change of the image. Similarly, by setting the minimum interval number between two peaks, the situation where the function judges multiple extreme values within the same period as peak points can be avoided. The emergence of this correction form can assist the model to accurately read the peak points of the waveform after each section of the image contour is unfolded, improve the sensitivity and capture accuracy of the peak points, avoid misreading multiple peak points, and not miss the peak positions within each waveform interval. In addition, when reading peak points, there are special cases of data that need to be corrected. In this case, although the minimum height and minimum interval number of the wave peaks can be set to extract peak points more accurately, when the draped image is unfolded at the peak point, the function will read the same peak point twice. To avoid this situation, a new judgment needs to be added to the function: for example, if the distance between the front and rear points and the edge point is less than 20 pixel points, then compare the sizes of the wave peak angles of the two points and retain the peak point with the larger angle. The corrected extraction of wave peak points is shown in the figure. After the wave peak point positioning is completed, the coordinates of the wave peak points are stored. After the peak points are read, the draped parameters of the fabric are obtained through the static draping coefficient calculation formula. The calculation method of the static draping coefficient is as follows: Among them, A1 is the projected area of the fabric, A2 is the area of the fabric tray, and A3 is the area of the fabric. In addition, wave number acquisition is also required. After the draped projection contour of the fabric is unfolded, there are wave peaks or trough numbers. Counting the number of wave peaks and troughs can obtain the wave number; confirmation of the wave peak angle, the wave peak angle refers to the angle between two adjacent wave peaks of the fabric projection contour; determining the wave peak amplitude according to the obtained wave peak situation, the wave peak amplitude is mainly determined as the distance from the wave peak point in the fabric draped projection to the edge of the fabric tray. After processing the draped image and counting all the parameters, it is input into the hidden layer of the multi-layer neural network model for training, and finally the bending hardness characteristics of the fabric are determined through the fabric draped image for the reference of relevant personnel.

[0028] In this embodiment, a neural network model is constructed and optimized. This model predicts the influence of fabric drape parameters on the control parameters for garment manufacturing. In the multi-layer neural network model, the input layer is set with 8 fabric drape parameters, and the output layer is 7 fabric control parameters. After training, the mean square error and the accuracy of the correlation coefficient of this model are further improved. Further, the model is optimized through correlation analysis. The values of 3 control parameters with relatively low correlation with the fabric drape parameters are determined as standard values. Then, a multi-layer neural network is established and the fabric drape parameters are input into the input layer. The fabric drape parameters include drape coefficient, wave number, wave peak amplitude, and wave peak angle. The output layer is 4 main control parameters related to bending with relatively strong correlation with the fabric drape parameters. At the same time, the number of hidden layers in the multi-layer neural network model is 1, and the number of hidden layer nodes is determined to be 12. Finally, the activation function is determined as the sigmoid function. After the above settings are completed, the accuracy of the mean square error and the correlation coefficient included in the network training results of the multi-layer network model in this embodiment is significantly improved. And the drape performance of real fabrics is tested and simulated. By pairing the real fabrics with the garment manufacturing results, there is no obvious difference in 3 groups of paired data of drape coefficient, wave number, and maximum wave peak amplitude, thus completing the judgment of the bending hardness and actual characteristics of various fabrics, and improving the efficiency of clothing manufacturing, the accuracy of garment manufacturing, and the yield rate of good products.

[0029] Embodiment 2: This embodiment includes a device for fabric evaluation. When obtaining the bending hardness of the fabric, it is necessary to first fix the fabric to be tested according to the cantilever method, and then use an image acquisition module such as a camera to take and collect the drape image during the bending hardness test of the fabric. Then, the obtained drape image is imported into the neural network module for processing. After the processing is completed, the neural network module exports the processing result of the drape image. Finally, the processing result will display the simulation test result of the fabric bending degree on the evaluation module.

[0030] Embodiment 3: The present invention also provides a computer-readable storage medium, which includes a fabric drape image processing program. When there is a fabric drape image to be processed, the fabric drape image processing program will be executed, and any computer device running the fabric drape image processing program will execute according to the established fabric drape image processing method. In this embodiment, the computer-readable storage medium can be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system, device, or component. In the present invention, the computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries the computer-readable program code. Such a propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer-readable signal medium can also be any computer-readable medium other than the computer-readable storage medium, and this computer-readable medium can send, propagate, or transmit a program for use by or in combination with an instruction execution system, device, or component. The program code contained on the computer-readable medium can be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0031] Embodiment 4: The present invention includes a computer device. The computer device in this embodiment includes a memory for storing a computer program; and a processor for executing the computer program stored in the memory. Relevant staff can use the computer device in this embodiment to perform a simulation evaluation of the fabric bending hardness.

Claims

1. A fabric evaluation method, characterized in that, The steps include S1: Establish a multi-layer neural network model to obtain the fabric drape image and perform parametric processing; S2: Extract the image contour of the fabric drape image and unfold it to complete the fabric evaluation.

2. The fabric evaluation method according to claim 1, characterized in that, In step S1, several fabric control parameters are preset in advance. After obtaining the parameters, the input layer, output layer, and hidden layer of the neural network are constructed. The number of layers and nodes of each layer of the input layer, output layer, and hidden layer, as well as the type of activation function, are set respectively. The actual photographed image of the fabric is used as the input vector of the multi-layer neural network, and the fabric bending hardness is used as the output vector of the multi-layer neural network to realize the mapping from the fabric image to the fabric drape parameters.

3. The fabric evaluation method according to claim 1, characterized in that, In step S2, after the multi-layer neural network model is established, parametric processing of the drape image is performed according to the fabric test image and model structure obtained by the multi-layer neural network model. Calculate the image area after parametric preprocessing of the drape image, extract the image contour of the drape image and unfold it, sample the peak points of the unfolded image, and obtain the drape parameters.

4. The fabric evaluation method according to claim 3, characterized in that In step S2, the parametric processing of the drape image includes at least one of the processing methods such as image preprocessing, image area calculation, image contour extraction, image contour unfolding, peak point extraction, and drape parameter extraction to obtain the drape index of the virtual fabric.

5. The fabric evaluation method according to claim 4, characterized in that, In step S2, the image preprocessing step includes converting the original fabric drape image into a grayscale image and taking the inverse color, and the inverse-colored image is segmented after filtering.

6. The fabric evaluation method according to claim 5, characterized in that In step S2, the image is converted into a binary image and the image area is calculated. When calculating the area, the pixel value of the points in the drape image area is characterized as 1, and the pixel value of the background is characterized as 0. The calculation method of the image area is: A = ∑ (x，y)∈R f(x, y), In the formula, R is the set of pixels in the drape image area, x and y are the two-dimensional coordinates of the drape image, and the number of statistical characterizations is the drape image area.

7. A method for fabric evaluation according to any one of claims 5 or 6, characterized in that, In step S2, after the above processing of the drape image, the following steps are further included: extracting and collecting the drape image contour, collecting the drape image contour, and extracting the internal edge formed by the internal contour holes of the drape image while extracting the drape external contour; unfolding the drape image contour, constructing a graphic circle with the center of the fabric drape image as the center, determining the radius and angle of the graphic circle, and unfolding and extending the image contour according to the radius and angle of the circle.

8. A method for fabric evaluation according to claim 3, characterized in that, In step S2, the drape parameters are obtained by calculating the static drape coefficient. The calculation method of the static drape coefficient is: Where A1 is the fabric projection area, A2 is the fabric tray area, and A3 is the fabric area.

9. A method for fabric evaluation according to claim 3, characterized in that In step S2, the peak point sampling step further includes: wave number collection. After the fabric drape projection contour is unfolded, there are peaks or valleys. Counting the number of peaks and valleys can obtain the wave number; confirmation of the peak angle. The peak angle refers to the angle between two adjacent peaks of the fabric projection contour; determining the peak amplitude according to the obtained peak situation. The peak amplitude mainly determines the distance from the peak point in the fabric drape projection to the edge of the fabric tray.

10. An apparatus for fabric evaluation applicable to any one of claims 1 to 9, comprising an image acquisition module for acquiring real-time images of fabric bending hardness tests; a neural network module for processing fabric draping images to obtain draping image processing results; and an evaluation module for displaying fabric curvature simulation test results, evaluating the fabric curvature and feeding back the results.

11. A storage medium applicable to any one of claims 1 to 9, comprising a draping image processing program which, when executed, causes the computer to execute according to a predetermined draping image processing method.

12. A computer device applicable to any one of claims 1 to 9, comprising a memory and a processor, where the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, characterized in that, When the computer program runs, it causes the processor to execute the fabric evaluation method according to any one of claims 1 to 8.

Images