Fabric appearance flatness intelligent visual rating device and method

Abstract

The invention discloses an intelligent visual rating device and method for fabric appearance flatness. The method comprises the following steps: S1, acquiring a fabric image; s2, preprocessing the fabric image; s3, camera calibration and image correction; s4, feature extraction and matching; s5, stereo matching and depth estimation; s6, performing three-dimensional reconstruction; s7, grading the appearance flatness of the fabric; the objective rating method for automatically processing the fabric depth image difference is constructed on the basis of quickly reconstructing the fabric appearance form based on a machine vision technology, so that the fabric flatness detection efficiency and accuracy are improved, and a more reliable and objective basis is provided for fabric quality evaluation.

Description

Technical Field

[0001] The invention belongs to the field of fabric detection, and in particular relates to an intelligent visual rating device and method for fabric appearance smoothness. Background Art

[0002] During the production, processing and use of flexible textile fabrics, they are easily wrinkled due to the influence of mechanical forces, detergents and environmental factors, which affects the aesthetics, comfort and durability of textile clothing. At present, the traditional method of detecting the smoothness of fabric appearance mainly relies on manual visual inspection, which is highly subjective and inefficient, and cannot meet the current product quality inspection needs. An intelligent, fast and objective detection method is urgently needed.

[0003] In the published patents, the image processing method mainly relies on image technology to evaluate the flatness of fabrics, covering the extraction, analysis and processing of image features. For example, the invention patent with application number CN201910325894.4 discloses a fabric flatness evaluation method based on the combination of three-dimensional images and two-dimensional image features. This method combines the two-dimensional information of the fabric with the three-dimensional information to obtain more fabric surface information. This combination avoids the shortcomings of single-dimensional images in characterizing the flatness of fabrics, and converts the flatness problem of the curved surface into a more easily solved problem of the degree of curvature of the curve. However, the applicability of this method is limited for fabrics with different textures and structures, and if the two-dimensional and three-dimensional data are incorrectly merged, inaccurate evaluation may result. The invention patent with application number CN201811165240.1 discloses an objective rating method for fabric flatness based on Fourier spectrum features, which can locate the frequency range of wrinkles recognized by the human eye, filter out information that the human eye cannot recognize, and is more in line with the human eye visual evaluation system. However, this method may ignore some tiny details that affect the quality of the fabric, resulting in the loss of visual information. In addition, a single frequency range selection may not fully reflect the complexity of different fabrics. The invention patent with application number CN201810613634.2 discloses an objective evaluation method for fabric smoothness based on four-side light source images. It has the characteristics of high detection accuracy, objectivity, stability, reproducibility and automation, but is easily affected by the external lighting environment. When faced with complex patterns, textures or fabric defects, uneven light reflection may occur, thus affecting the evaluation results. The invention patent with application number CN201510710539.0 proposes a method for testing and evaluating the smoothness of clothing. It uses computer image processing technology to test and evaluate the smoothness of clothing. It is suitable for uneven detection caused by wrinkling due to movement, but in actual wearing, rapidly changing wrinkles may cause inaccurate detection.

[0004] Machine learning methods focus on analyzing and rating fabric features through algorithm and model training. For example, the invention patent with application number CN201510020830.5 discloses a fabric appearance flatness evaluation method based on sparse coding, which realizes automatic rating under unsupervised state, but its effect depends on the richness of the selected base dictionary and is highly sensitive to data quality. There is a risk of poor model generalization ability and it may not perform well on new fabrics. The invention patent with application number CN201610930156.9 discloses a fabric flatness customer evaluation method and device based on unsupervised machine learning, which performs fabric rating through feature extraction and clustering to reduce manual errors, but the limitations of feature selection and the uncertainty of clustering results may affect the accuracy and stability of the evaluation.

[0005] In terms of physical detection devices, the utility model patent with application number CN202121643103.1 proposes a heat-set knitted fabric flatness control detection device, which solves the problem of real-time sample comparison, but its application scope is limited. The invention patent with application number CN201510710673.0 discloses a fabric post-washing twisting flatness test device and method, filling the gap in hand-washing and wringing flatness detection, but it is only applicable to fabrics under specific conditions and its application scope is limited.

[0006] Although the patents that have been published so far involve methods for detecting fabric flatness, they are limited in scope of application. They are usually only applicable to specific types of fabric samples, lack specific technical discussions on fabric flatness under interference from color, texture, complex patterns, etc., and cannot meet the diverse needs of the market. In addition, the existing methods are easily affected by the external lighting environment, resulting in insufficient ability to capture details. Therefore, it is urgent to develop an objective detection method for fabric appearance flatness that is highly applicable, has high detection efficiency, and produces accurate and reliable evaluation results. Summary of the invention

[0007] The purpose of the present invention is to provide an intelligent visual rating device and method for fabric appearance smoothness. Based on machine vision technology, on the basis of rapid reconstruction of fabric surface morphology, an objective rating method for automatically processing fabric depth image differences is constructed to improve the efficiency and accuracy of fabric smoothness detection, and provide a more reliable and objective basis for fabric quality assessment.

[0008] In order to solve the above technical problems, the following technical solutions are adopted:

[0009] An intelligent visual rating method for fabric appearance smoothness, characterized by comprising the following steps:

[0010] S1: fabric image acquisition;

[0011] S2: fabric image preprocessing;

[0012] S3: Camera calibration and image correction: The checkerboard calibration method is used to obtain the intrinsic and extrinsic parameters of the binocular stereo camera and establish the geometric model of camera imaging. Through calibration, the translation vector and rotation vector between the left and right cameras are obtained, and epipolar geometry correction is performed based on this to eliminate image distortion caused by lens distortion, ensure epipolar constraints between images, and provide basic data for 3D reconstruction.

[0013] S4: Feature extraction and matching: Based on the deep learning method, a global convolutional neural network is used as a feature extractor to extract feature points or feature regions for subsequent matching from the left and right images of the fabric. Subsequently, the correspondence between the feature points or feature regions in the left and right images of the fabric is established, and the similarity or distance measurement between the features is used for matching, and a feature correlation pyramid is constructed to capture the correlation between features at different scales. The information in the feature correlation pyramid is used to calculate the matching cost between the features of the left and right images of the fabric. At the same time, a competition mechanism is introduced to compare the matching degree between different features and select the best match.

[0014] S5: Stereo matching and depth estimation: Stereo matching is to calculate the horizontal displacement between corresponding points in the left and right images of the fabric, i.e., the disparity value, based on the result of feature matching. Then, in the depth estimation stage, the calculated disparity value is converted into a depth value using the camera parameters and the disparity-depth conversion formula, thereby constructing a depth map.

[0015] S6: 3D reconstruction: Generate the 3D point cloud structure of the fabric based on the depth map and camera parameters, and then perform point cloud processing, triangular mesh model generation, post-processing, visualization analysis and display;

[0016] S7: Fabric appearance smoothness rating: Based on the reconstructed three-dimensional point cloud data of the fabric, the three-dimensional spatial morphology of the fabric surface is deeply analyzed, and then the fabric appearance smoothness is graded.

[0017] The intelligent visual rating method for fabric appearance smoothness according to claim 1 is characterized in that: step S1 comprises:

[0018] (1) Fabric sample cutting: Cut fabric samples of equal size from different parts of the same fabric;

[0019] (2) Fabric sample processing: the cut fabric sample is subjected to temperature adjustment and then subjected to one or more combined treatments of washing, drying and ironing, and the treated fabric sample is trimmed;

[0020] (3) Image acquisition: Use an image acquisition device to photograph the fabric sample to obtain a fabric image.

[0021] After optimization, the step S2 includes effective area cropping, grayscale processing, image enhancement, lighting balance and noise reduction processing.

[0022] After optimization, the step S3 includes:

[0023] (1) Camera calibration: The binocular stereo camera is calibrated using the chessboard calibration method, where the chessboard calibration board consists of black and white grids. Under the premise of uniform light and stable installation of the binocular stereo camera, the chessboard calibration board images are captured at different positions and angles to obtain left and right view data. Then, the inner corner points on the chessboard calibration board are extracted. Based on the initially extracted inner corner point information, the sub-pixel information is extracted using the sub-pixel refinement algorithm to reduce the camera calibration deviation. Finally, the binocular stereo camera calibration is completed and the camera's intrinsic and extrinsic parameters are obtained.

[0024] (2) Image correction: Using the camera intrinsic and extrinsic parameters calculated in step (1), perform intrinsic correction and epipolar geometry correction on the image to eliminate the image distortion caused by camera lens distortion. Assume that a point P in Euclidean space has coordinates P1 = (x1, y1, z1) in the two coordinate systems. T and P2=(x2,y2,z2) T , then the transformation formula is:

[0025] P2=RP1+t (1)

[0026] In formula (1), R is the rotation matrix, which describes the rotation relationship between the two camera coordinate systems; t is the translation vector, which describes the translation relationship between the two camera coordinate systems; P1 and P2 are the pixel coordinates of the projection points of point P on the two camera image planes, respectively.

[0027] The intelligent visual rating method for fabric appearance smoothness according to claim 1 is characterized in that: step S4 comprises:

[0028] (1) Feature extractor construction: Use the pre-trained convolutional neural network to perform deep feature extraction on the input left and right images of the fabric;

[0029] (2) Feature correlation pyramid construction: The extracted feature points or feature regions are used to construct a feature correlation pyramid; in each layer of the pyramid, the similarity or distance measurement between the features is used for matching, thereby establishing the correspondence between the feature points or feature regions in the left and right images of the fabric;

[0030] (3) Application of gradient-related region update operator: In the process of feature matching, the gradient-related region update operator is introduced to optimize the matching results in an iterative manner. By analyzing the gradient information around the feature point, the operator can adjust the position and weight of the feature point.

[0031] (4) Matching cost calculation and competition mechanism application: The information in the feature correlation pyramid is used to calculate the matching cost between the left and right image features of the fabric. At the same time, a competition mechanism is introduced to compare the matching degree between different features, and the optimal match is selected by setting a threshold or sorting method.

[0032] After optimization, the step S5 comprises:

[0033] (1) Stereo matching: Based on the result of feature matching, the horizontal displacement between corresponding points in the left and right images of the fabric, i.e., the disparity value, is calculated. A stereo matching algorithm that integrates semi-global matching, adaptive weight aggregation, and cross-based cost aggregation modules is used to optimize the preliminary matching results and obtain a more accurate disparity value.

[0034] (2) Depth estimation: Based on the camera parameters and disparity values, the depth value of each pixel is calculated using the disparity-depth conversion formula (2):

[0035]

[0036] In formula (2), Z is the depth value; f is the focal length of the camera; B is the distance between p and p′; d is the disparity value, which refers to the difference in pixel positions between the left and right images of the binocular stereo camera. The X coordinates of point p and point p′ in the corresponding phase plane are x and l and x r , define the disparity value d = x l -x r ;

[0037] (3) Optimize the depth map: By comparing the depth values ​​of corresponding points in the left and right images of the fabric, inconsistent matching points are eliminated. For points whose differences exceed a certain threshold, they are removed from the depth map.

[0038] After optimization, the step S6 comprises:

[0039] (1) 3D point cloud generation: Using the depth value of each pixel in the depth image and combining it with the camera parameters, the pixel is converted from the 2D image space to the 3D world space to generate a 3D point cloud.

[0040] (2) Point cloud processing: Smooth the generated 3D point cloud to eliminate noise and burrs; use statistical filtering and radius filtering to remove isolated points and noise points in the point cloud;

[0041] (3) Triangular mesh generation: Using triangular mesh generation algorithms such as Poisson surface reconstruction, the processed point cloud is triangulated to generate a continuous triangular mesh model;

[0042] (4) Post-processing: Optimize and adjust the initially generated triangular mesh model, map the texture information in the original fabric image to the triangular mesh model to enhance the realism of the model; optimize the edges and vertices of the triangular mesh to improve the fineness and smoothness of the model;

[0043] (5) Visual analysis and display: Use 3D visualization software to visualize and analyze the generated triangular mesh model. Through the intuitive 3D model, the microscopic morphology and undulations of the fabric can be directly observed.

[0044] After optimization, the step S8 comprises:

[0045] (1) Data input and preprocessing: Input the fabric point cloud data of the new sample into the trained triangular mesh model; the input data should be in the same format as the training data, including 3D geometric features and preprocessed feature values;

[0046] (2) Grade prediction: The triangular mesh model calculates the flatness grade of the fabric based on the input point cloud data through a deep learning algorithm and outputs a rating result from 1 to 5. During the prediction process, the model uses the relationship between the features and labels learned in the previous training to derive the corresponding flatness grade.

[0047] (3) Result comparison and verification: The flatness grade predicted by the triangular mesh model is compared with the sample flatness grade obtained through subjective evaluation. At the same time, the real flatness data is used as a reference to verify whether the prediction of the triangular mesh model is consistent with the actual situation.

[0048] (4) Accuracy optimization: In the feature extraction stage, high-resolution point cloud data and refined geometric feature analysis are used to capture the subtle height changes and wrinkle features on the fabric surface. The support vector machine algorithm is used for flatness rating, and during the rating, the cross-validation technology is used to verify the triangular mesh model multiple times.

[0049] (5) Result output: The triangular mesh model will output the flatness rating of each sample and its corresponding geometric feature data.

[0050] An intelligent visual rating device for fabric appearance smoothness, characterized in that it includes a three-dimensional frame, a binocular stereo camera and an LED light source, the bottom of the three-dimensional frame is a fabric loading platform, the top of the three-dimensional frame is a top plate, the fabric loading platform is provided with a cloth loading platform, and the cloth loading platform is detachable; the LED light source is arranged on both sides of the binocular stereo camera, the fabric loading platform is connected to a bottom stretching and extension component, the bottom stretching and extension component is used to extend the fabric loading platform, the bottom stretching and extension component is equipped with a homogeneous filling plate, and the homogeneous filling plate is filled in the bottom stretching and extension component after expansion; the top plate is connected to a top stretching and extension component, the top stretching and extension component is used to extend the top plate, the top stretching and extension component is equipped with a lightweight filling plate, and the lightweight filling plate is filled in the top stretching and extension component after expansion.

[0051] After optimization, the top plate is provided with a built-in magnetic ring belt, and the bottom stretch expansion component is provided with an external magnetic belt. The built-in magnetic ring belt and the external magnetic belt are used to fix the magnetic flexible curtain, and the magnetic flexible curtain is opaque.

[0052] The above technical solution has the following beneficial effects:

[0053] 1. The present invention is not affected by factors such as fabric color, texture, and complex patterns, and is suitable for flatness detection of various types of fabrics. The detection device can also expand the characteristics of the detection space, so that it can be expanded to the flatness detection of the appearance of ready-made clothes, and has wide applicability and application prospects.

[0054] 2. By optimizing the image processing algorithm and ensuring the stability of lighting at the physical level, the present invention effectively reduces the impact of the external lighting environment on the detection results, ensuring that stable and accurate detection results can be obtained under different lighting conditions.

[0055] 3. Through the trained machine learning model, the present invention can achieve objective and accurate rating of fabric point cloud data, with a rating accuracy of 0.1, providing a scientific basis for fabric quality assessment.

[0056] 4. As an efficient and accurate detection method, machine vision technology has shown good application potential in the evaluation of fabric appearance flatness. Using machine vision technology, the fabric surface can be comprehensively scanned and analyzed, and the characteristic information of surface flatness, such as the height and density of wrinkles, can be automatically identified and extracted. The key to this technology is to quickly extract and evaluate fabric surface features through image processing and analysis algorithms. Therefore, the present invention proposes an automatic detection method for fabric appearance flatness based on machine vision, which can efficiently and accurately calculate the flatness grade of the fabric, providing a reliable basis for fabric quality evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] The present invention will be further described below in conjunction with the accompanying drawings:

[0058] Figure 1 The figure is a flow chart of the rating method of the present invention.

[0059] Figure 2 It is a schematic diagram of the structure of the rating device of the present invention.

[0060] Figure 3 This is a schematic diagram of the arrangement of the magnetic flexible curtain;

[0061] Figure 4 A schematic diagram of a homogeneous filler plate or a lightweight filler plate;

[0062] Figure 5 Train the network architecture for the triangular mesh model.

[0063] The accompanying drawings are marked as follows: fabric loading platform 1, fabric loading platform 2, bottom stretching and extension component 3, top plate 4, binocular stereo camera 5, LED light source, adjustable camera bracket 7, shockproof pad 8, stereo frame 9, top stretching and extension component 10, built-in magnetic ring belt 11, external magnetic belt 12, magnetic flexible curtain 13. DETAILED DESCRIPTION

[0064] The present invention aims to provide an intelligent visual rating device and method for fabric appearance smoothness. Based on machine vision technology, on the basis of rapid reconstruction of fabric appearance, an objective rating method for automatically processing fabric depth image differences is constructed to improve the efficiency and accuracy of fabric smoothness detection and provide a more reliable and objective basis for fabric quality assessment. The technical solution of the present invention is described in detail below in conjunction with specific embodiments:

[0065] Example 1

[0066] like Figure 2As shown, the present invention designs an intelligent visual rating device for fabric appearance flatness, which is used for fabric image acquisition. The device includes a three-dimensional frame 9, a binocular stereo camera 5 and an LED light source 6. The three sides of the three-dimensional frame 9 are open. As the basic structure of the rating device, the bottom of the three-dimensional frame 9 is a fabric stage 1, and the top of the three-dimensional frame 9 is a top plate 4. The fabric stage 1 is made of a gray metal plate to ensure that the fabric remains flat and does not deform during the image acquisition process. The fabric stage 1 is connected to a bottom stretching and expansion component 3, which can be stretched to a required size according to the size of the fabric sample. After the bottom stretching and expansion component 3 is expanded, a through hole is formed inside. In order to ensure that the fabric stage 1 is flat after expansion, it is equipped with a homogeneous filling plate, whose structure matches the through hole and is used to fill the through hole, thereby forming a complete, seamless horizontal plane, eliminating the layer difference generated after expansion, and the puzzle-like design of the homogeneous filling plate ensures that the bottom of the rating device does not leak light. The plane expansion function allows the device and rating method to be expanded from testing fabric samples to rating the flatness of finished garment appearance.

[0067] The fabric stage 1 is provided with a fabric stage 2, which is used to place the fabric sample and is detachable and can be adaptively replaced according to the size of the fabric sample. A shockproof pad 8 is provided at the bottom of the fabric stage 1 to reduce the impact of external environmental vibration on image quality, ensure stability during shooting, and improve the reliability of image acquisition.

[0068] The top plate 4 is connected to a top stretching and expansion component 10, which has the same structure and function as the bottom stretching and expansion component 3. Similarly, after the top stretching and expansion component 10 is expanded, a through hole is formed inside it. In order to ensure that the fabric loading platform 1 is flat after being unfolded, it is equipped with a lightweight filling plate, whose structure matches the through hole and is used to fill the through hole. Compared with the same filling plate, the lightweight filling plate is lighter and reduces the load-bearing capacity of the top structure.

[0069] The binocular stereo camera 5 is the core imaging device, which is installed on the stereo frame 9 through an adjustable camera bracket 7 and is located on the top of the stereo frame 9; the binocular stereo camera 5 has a high resolution of more than 5 million pixels and is equipped with automatic focus and exposure adjustment functions to adapt to image acquisition under different lighting conditions. At the same time, before the image information is collected, the viewing angle and position between the two cameras must be accurately adjusted to ensure the consistency of the depth information of the image they acquire. The binocular stereo camera 5 is equipped with LED light sources 6 with adjustable brightness on both sides, which are used to provide uniform lighting, reduce the impact of shadows and reflections on image quality, and avoid the loss of image information caused by uneven lighting. The height and angle of the adjustable camera bracket 7 can be adjusted according to experimental needs.

[0070] The binocular stereo camera 5 is connected to the computer via a USB or Ethernet cable to facilitate real-time transmission and storage of images. The computer is equipped with high-performance image processing software, which supports real-time image analysis and processing functions, and provides support for subsequent image preprocessing, correction and three-dimensional reconstruction.

[0071] The top plate 4 is provided with a built-in magnetic belt 11, and the bottom stretching and expansion component 3 is provided with an external magnetic belt 12. The built-in magnetic belt 11 and the external magnetic belt 12 are used to fix the magnetic flexible curtain 13, which is black and opaque. The built-in magnetic belt 11, the external magnetic belt 12 and the magnetic flexible curtain 13 are used in combination to form an opaque environment, ensuring that the test light source is clean and unique, and avoiding the influence of the external light environment. The built-in magnetic belt 11 is connected to the top plate 4, and it is attracted to the magnetic area on the upper part of the magnetic flexible curtain 13, while the external magnetic belt 12 is adsorbed on the bottom stretching and expansion component 3, connected to the closed magnetic flexible curtain 13, to achieve the creation of an opaque space and reduce the influence of the external light environment. Considering that the perimeter of the three sides changes after the space is expanded, the length of the black magnetic flexible curtain 13 is slightly longer than the perimeter of the device in the fully expanded state. During use, the excess part can be magnetically attracted to the three-dimensional frame 9.

[0072] Example 2

[0073] The present invention provides an intelligent visual rating device and method for fabric appearance smoothness, comprising the following steps:

[0074] S1: Fabric image acquisition:

[0075] Since fabric samples usually show different degrees of undulations and wrinkles after washing, drying or ironing, this provides typical characteristics for the determination of fabric appearance flatness. According to relevant standards or user requirements, three pieces of fabric samples of the same size after washing, drying or ironing are randomly cut from different areas of a piece of fabric. After the above treatment, the image acquisition device of Example 1 is used to collect standard fabric appearance morphology images of the three samples. The specific steps are as follows:

[0076] (1) Fabric sample cutting: According to relevant standards or user requirements, three fabric samples of equal size are cut from different parts of the same fabric. Ensure that the sample sizes are consistent to reduce the impact of human errors on the test results.

[0077] (2) Fabric sample processing: The cut samples are subjected to temperature adjustment for 15 hours or more, and then washed, dried, ironed, or subjected to one or more combination of processes in accordance with relevant standards, and the treated fabric samples are trimmed.

[0078] a. Washing: Place the fabric sample in a washing machine and wash it with clean water or detergent to remove possible impurities. After washing, dry it naturally to create natural wrinkles.

[0079] b. Ironing: Use an iron to iron to different levels of wrinkle effects. The temperature and pressure should be controlled during ironing to form repeatable wrinkle characteristics.

[0080] c. Trimming: Some fabrics may have burrs after washing, so the excess burrs need to be trimmed to ensure that the sample remains intact and has no excess parts.

[0081] (3) Image acquisition: The fabric samples processed in steps (1) and (2) are photographed using the device of Example 1 to ensure that the surface morphology of each sample is fully recorded, thereby providing important visual data support for the evaluation of the smoothness of the fabric appearance.

[0082] S2: fabric image preprocessing;

[0083] In order to provide a reliable basis for subsequent image analysis and fabric quality assessment, the obtained fabric image is subjected to effective area cropping, grayscale processing, image enhancement combined with light balance to highlight detail information, and then smoothing to reduce noise interference. The specific steps are as follows:

[0084] (1) The effective area is cut to

[0085] Focus on analyzing the information on the fabric surface, avoid external interference, use edge detection technology, automatically identify and crop the effective area of ​​the fabric image, remove the redundant parts, and ensure that the image information is concise and effective.

[0086] (2) Grayscale processing

[0087] In order to simplify the image information and reduce the computational complexity, the collected fabric images are grayed out. By removing the color information and converting it into a grayscale image containing only brightness information, it is beneficial to the stability and accuracy of the subsequent processing steps.

[0088] (3) Image enhancement

[0089] The adaptive gamma correction method in the local contrast enhancement technology is used for image enhancement to accurately reflect the texture and wrinkle information of the fabric surface. Adaptive gamma correction dynamically adjusts the gamma value according to the local brightness information of the image to enhance the contrast of the image. In areas with lower brightness, the gamma value is reduced to enhance the dark details; in areas with higher brightness, the gamma value is increased to compress the bright details. This step not only retains the texture and wrinkle information of the fabric surface, but also makes the overall contrast of the image more uniform, highlighting the surface information of the fabric after processing.

[0090] (4) Light balance

[0091] An image enhancement method based on Retinex theory is adopted to simulate the adaptability of the human visual system to lighting changes. The lighting is adjusted by calculating the relationship between the brightness of each pixel in the image and the brightness of the surrounding pixels, eliminating the influence of lighting changes on fabric feature extraction, so that the image can maintain consistent brightness and contrast under different lighting conditions, thereby more accurately reflecting the characteristics of the fabric.

[0092] (5) Noise reduction processing

[0093] Gaussian filter combined with bilateral filter is used for noise reduction to remove random noise in the image. The introduction of bilateral filter significantly enhances the edge preservation ability and avoids the edge blurring problem that may be caused by Gaussian filter. This allows the processed image to clearly display the edge and texture information of the image while removing noise.

[0094] S3: Camera calibration and image correction:

[0095] The chessboard calibration method is used to obtain the intrinsic and extrinsic parameters of the binocular stereo camera and establish the geometric model of camera imaging. Through calibration, the translation vector and rotation vector between the left and right cameras are obtained, and epipolar geometry correction is performed based on this to eliminate image distortion caused by lens distortion, ensure epipolar constraints between images, and provide basic data for 3D reconstruction. The specific steps are as follows:

[0096] (1) Camera calibration: The binocular stereo camera is calibrated using the chessboard calibration method, where the chessboard calibration plate consists of black and white grids. To ensure that the camera parameters are accurate and effective, the chessboard calibration plate images are taken at different positions and angles under the premise of uniform light and stable installation of the binocular stereo camera, and a total of more than 7 left and right view data are obtained. Then, the inner corner points on the chessboard calibration plate are extracted, and the sub-pixel information is further extracted based on the initially extracted inner corner point information using the sub-pixel refinement algorithm, thereby reducing the camera calibration deviation. Finally, the binocular stereo camera calibration is completed and the camera's intrinsic and extrinsic parameters are obtained, including distortion coefficient, focal length, principal point position, etc.

[0097] (2) Image correction: Using the camera intrinsic and extrinsic parameters calculated in step (1), perform intrinsic correction and epipolar geometry correction on the image to eliminate the image distortion caused by camera lens distortion. Assume that a point P in Euclidean space has coordinates P1 = (x1, y1, z1) in the two coordinate systems. T and P2=(x2,y2,z2) T , then the transformation formula is:

[0098] P2=RP1+t (1)

[0099] In formula (1), R is the rotation matrix, which describes the rotation relationship between the two camera coordinate systems; t is the translation vector, which describes the translation relationship between the two camera coordinate systems; P1 and P2 are the pixel coordinates of the projection points of point P on the two camera image planes, respectively.

[0100] S4: Feature extraction and matching:

[0101] Based on the deep learning method, a global convolutional neural network is used as a feature extractor to extract feature points or feature regions for subsequent matching from the left and right images of the fabric; then, the correspondence between the feature points or feature regions in the left and right images of the fabric is established, and the similarity or distance measurement between the features is used for matching, and a feature correlation pyramid is constructed to capture the correlation between features at different scales; the information in the feature correlation pyramid is used to calculate the matching cost between the features of the left and right images of the fabric; at the same time, a competition mechanism is introduced to compare the matching degree between different features and select the optimal match; this step includes a feature extractor module, a feature correlation pyramid module, a gradient correlation region update operator module, etc. The specific steps are as follows:

[0102] (1) Feature extractor construction: Using the pre-trained convolutional neural network, deep feature extraction is performed on the input left and right images. This module can capture local details and global structural information in the image and generate high-dimensional feature representation.

[0103] (2) Feature correlation pyramid construction: The extracted feature points or feature regions are used to construct feature correlation pyramids, which can enhance the robustness of matching by capturing the correlation between features at different scales. In each layer of the pyramid, the similarity or distance metric between features is used for matching, thereby establishing the correspondence between the feature points or feature regions in the left and right images.

[0104] (3) Application of gradient-correlated region update operator: In the process of feature matching, the gradient-correlated region update operator is introduced to optimize the matching results in an iterative manner. By analyzing the gradient information around the feature points, the operator can adjust the position and weight of the feature points to improve the matching accuracy.

[0105] (4) Matching cost calculation and competition mechanism application: The information in the feature correlation pyramid is used to calculate the matching cost between the left and right image features of the fabric. At the same time, a competition mechanism is introduced to compare the matching degree between different features. The optimal match is selected by setting thresholds or sorting methods, ensuring the stability and reliability of the matching results and reducing the possibility of mismatching.

[0106] S5: Stereo matching and depth estimation:

[0107] Stereo matching aims to accurately calculate the horizontal displacement, i.e., the disparity value, between corresponding points in the left and right images based on the results of feature matching. To achieve this goal, a multi-module fusion algorithm is used to improve the matching accuracy, forming a stereo matching algorithm that integrates semi-global matching, adaptive weight aggregation, and cross-based cost aggregation modules. Subsequently, in the depth estimation stage, the calculated disparity value is converted into a depth value using the camera parameters and the disparity-depth conversion formula to construct a depth map. To improve the accuracy of the depth map, a left-right consistency check is performed to eliminate inconsistent matching points. The specific steps are as follows:

[0108] (1) Stereo matching: Based on the result of feature matching, the horizontal displacement between corresponding points in the left and right images of the fabric, i.e., the disparity value, is calculated. A stereo matching algorithm that integrates semi-global matching (SGM), adaptive weight aggregation, and cross-based cost aggregation modules is used to optimize the preliminary matching results and obtain a more accurate disparity value.

[0109] (2) Depth estimation: Based on the camera parameters and disparity values, the depth value of each pixel is calculated using the disparity-depth conversion formula (2):

[0110]

[0111] In formula (2), Z is the depth value; f is the focal length of the camera; B is the distance between p and p′; d is the disparity value, which refers to the difference in pixel positions between the left and right images of the binocular stereo camera. The X coordinates of point p and point p′ in the corresponding phase plane are x and l and x r , define the disparity value d = x l -x r ;

[0112] (3) Optimizing the depth map: By comparing the depth values ​​of corresponding points in the left and right images of the fabric, inconsistent matching points are eliminated to improve the accuracy of the depth map. That is, for points whose differences exceed a certain threshold, they are removed from the depth map.

[0113] S6: 3D reconstruction:

[0114] Based on the depth map and camera parameters, the 3D point cloud structure of the fabric is generated, and then the point cloud processing software is used to smooth and denoise the generated 3D point cloud structure to improve its quality and visualization effect. This step involves a 3D reconstruction module that combines depth map conversion, triangular mesh generation and post-processing technology. The specific steps are as follows:

[0115] (1) 3D point cloud generation: Using the depth value Z of each pixel in the depth image and combining it with camera parameters, such as the internal and external parameter matrices, the pixel is converted from the 2D image space to the 3D world space to generate a 3D point cloud.

[0116] (2) Point cloud processing: In order to extract useful information from the original data, perform preliminary processing and optimization on it, and smooth the generated 3D point cloud to eliminate noise and burrs. Statistical filtering, radius filtering and other methods are used to remove isolated points and noise points in the point cloud to improve the quality of the point cloud.

[0117] (3) Triangular mesh generation: Poisson surface reconstruction and other triangular mesh generation algorithms are used to triangulate the processed point cloud and generate a continuous triangular mesh model to facilitate subsequent visualization analysis.

[0118] (4) Post-processing: Optimize and adjust the initially generated triangular mesh model to improve its quality and visualization effect. Map the texture information in the original fabric image to the triangular mesh model to enhance the realism of the model; optimize the edges, vertices and other details of the triangular mesh to improve the fineness and smoothness of the model;

[0119] (5) Visual analysis and display: The generated triangular mesh model is visualized and analyzed using 3D visualization software such as CloudCompare. Through the intuitive 3D model, the microscopic morphology and undulations of the fabric can be directly observed, providing strong support for the rating of the fabric's appearance smoothness.

[0120] S7: Fabric appearance smoothness rating: Based on the reconstructed fabric 3D point cloud data, the 3D spatial morphology of the fabric surface is deeply analyzed, and then the fabric appearance smoothness is graded. The present invention combines the geometric features such as the height difference of the fabric surface undulations and the density of wrinkles, and uses a machine learning algorithm to classify the fabric smoothness into 1 to 5 levels, and the rating accuracy can reach 0.1 level. The specific steps are as follows:

[0121] (1) Data input and preprocessing: Input the fabric point cloud data of the new fabric sample into the trained triangular mesh model; the input data should be in the same format as the training data, including the 3D geometric features and preprocessed feature values;

[0122] (2) Grade prediction: The triangular mesh model calculates the flatness grade of the fabric based on the input point cloud data through a deep learning algorithm and outputs a rating result from 1 to 5. During the prediction process, the model uses the relationship between the features and labels learned in the previous training to derive the corresponding flatness grade.

[0123] (3) Result comparison and verification: The flatness grade predicted by the triangular mesh model is compared with the sample flatness grade obtained through subjective evaluation. At the same time, the real flatness data is used as a reference to verify whether the prediction of the triangular mesh model is consistent with the actual situation.

[0124] (4) Accuracy optimization: In order to achieve a flatness rating accuracy of 0.1, in the feature extraction stage, high-resolution point cloud data and refined geometric feature analysis are used to capture the subtle height changes and wrinkle features on the fabric surface; the support vector machine (SVM) algorithm is used for flatness rating, and during the rating, the cross-validation technology is used to verify the triangular mesh model multiple times to ensure its stability and accuracy.

[0125] (5) Result output: The triangular mesh model will output the flatness rating of each sample and its corresponding geometric feature data to ensure the accuracy and reliability of the results.

[0126] The above are only specific embodiments of the present invention, but the technical features of the present invention are not limited thereto. Any simple changes, equivalent replacements or modifications made based on the present invention to solve basically the same technical problems and achieve basically the same technical effects are all included in the protection scope of the present invention.

Claims

1. An intelligent visual rating method for fabric appearance smoothness, characterized in that The steps include: S1: fabric image acquisition; S2: fabric image preprocessing; S3: Camera calibration and image correction: The checkerboard calibration method is used to obtain the intrinsic and extrinsic parameters of the binocular stereo camera and establish the geometric model of camera imaging. Through calibration, the translation vector and rotation vector between the left and right cameras are obtained, and epipolar geometry correction is performed based on this to eliminate image distortion caused by lens distortion, ensure epipolar constraints between images, and provide basic data for 3D reconstruction. S4: Feature extraction and matching: Based on the deep learning method, a global convolutional neural network is used as a feature extractor to extract feature points or feature areas for subsequent matching from the left and right image pairs of the fabric; Subsequently, the correspondence between the feature points or feature regions in the left and right images of the fabric is established, and the similarity or distance metric between the features is used for matching, and a feature correlation pyramid is constructed to capture the correlation between the features at different scales; The information in the feature correlation pyramid is used to calculate the matching cost between the left and right image features of the fabric. At the same time, a competition mechanism is introduced to compare the matching degree between different features and select the best match. S5: Stereo matching and depth estimation: Stereo matching is to calculate the horizontal displacement between corresponding points in the left and right images of the fabric, i.e., the disparity value, based on the result of feature matching; Then, in the depth estimation stage, the calculated disparity values ​​are converted into depth values ​​using camera parameters and the disparity-depth conversion formula, thereby constructing a depth map; S6: 3D reconstruction: Generate the 3D point cloud structure of the fabric based on the depth map and camera parameters, and then perform point cloud processing, triangular mesh model generation, post-processing, visualization analysis and display; S7: Fabric appearance smoothness rating: Based on the reconstructed three-dimensional point cloud data of the fabric, the three-dimensional spatial morphology of the fabric surface is deeply analyzed, and then the fabric appearance smoothness is graded.

2. The intelligent visual rating method for fabric appearance smoothness according to claim 1, characterized in that: The step S1 comprises: (1) Fabric sample cutting: Cut fabric samples of equal size from different parts of the same fabric; (2) Fabric sample processing: the cut fabric sample is subjected to temperature adjustment and then subjected to one or more combined treatments of washing, drying and ironing, and the treated fabric sample is trimmed; (3) Image acquisition: Use an image acquisition device to photograph the fabric sample to obtain a fabric image.

3. The intelligent visual rating method for fabric appearance smoothness according to claim 1, characterized in that: The step S2 includes effective area cropping, grayscale processing, image enhancement, lighting balance and noise reduction processing.

4. The intelligent visual rating method for fabric appearance smoothness according to claim 1, characterized in that: The step S3 comprises: (1) Camera calibration: The binocular stereo camera is calibrated using the chessboard calibration method, where the chessboard calibration board consists of black and white grids. Under the premise of uniform light and stable installation of the binocular stereo camera, the chessboard calibration board images are captured at different positions and angles to obtain left and right view data. Then, the inner corner points on the chessboard calibration board are extracted. Based on the initially extracted inner corner point information, the sub-pixel information is extracted using the sub-pixel refinement algorithm to reduce the camera calibration deviation. Finally, the binocular stereo camera calibration is completed and the camera's intrinsic and extrinsic parameters are obtained. (2) Image correction: Using the camera intrinsic and extrinsic parameters calculated in step (1), perform intrinsic correction and epipolar geometry correction on the image to eliminate the image distortion caused by camera lens distortion. Assume that a point P in Euclidean space has coordinates P1 = (x1, y1, z1) in the two coordinate systems. T and P2=(x2,y2,z2) T , then the transformation formula is: P2=RP1+t (1) In formula (1), R is the rotation matrix, which describes the rotation relationship between the two camera coordinate systems; t is the translation vector, which describes the translation relationship between the two camera coordinate systems; P1 and P2 are the pixel coordinates of the projection points of point P on the two camera image planes, respectively.

5. The intelligent visual rating method for fabric appearance smoothness according to claim 1, characterized in that: The step S4 comprises: (1) Feature extractor construction: Use the pre-trained convolutional neural network to perform deep feature extraction on the input left and right images of the fabric; (2) Feature correlation pyramid construction: The extracted feature points or feature regions are used to construct a feature correlation pyramid; in each layer of the pyramid, the similarity or distance measurement between the features is used for matching, thereby establishing the correspondence between the feature points or feature regions in the left and right images of the fabric; (3) Application of gradient-related region update operator: In the process of feature matching, the gradient-related region update operator is introduced to optimize the matching results in an iterative manner. By analyzing the gradient information around the feature point, the operator can adjust the position and weight of the feature point. (4) Matching cost calculation and competition mechanism application: The matching cost between the left and right image features of the fabric is calculated using the information in the feature correlation pyramid. At the same time, a competition mechanism is introduced to compare the matching degree between different features, and the optimal match is selected by setting a threshold or sorting method.

6. The intelligent visual rating method for fabric appearance smoothness according to claim 1, characterized in that: The step S5 comprises: (1) Stereo matching: Based on the result of feature matching, the horizontal displacement between corresponding points in the left and right images of the fabric, i.e., the disparity value, is calculated. A stereo matching algorithm that integrates semi-global matching, adaptive weight aggregation, and cross-based cost aggregation modules is used to optimize the preliminary matching results and obtain a more accurate disparity value. (2) Depth estimation: Based on the camera parameters and disparity values, the depth value of each pixel is calculated using the disparity-depth conversion formula (2): In formula (2), Z is the depth value; f is the focal length of the camera; B is the distance between p and p′; d is the disparity value, which refers to the difference in pixel positions between the left and right images of the binocular stereo camera. The X coordinates of point p and point p′ in the corresponding phase plane are x and l and x r , define the disparity value d = x l -x r ; (3) Optimize the depth map: By comparing the depth values ​​of corresponding points in the left and right images of the fabric, inconsistent matching points are eliminated. For points whose differences exceed a certain threshold, they are removed from the depth map.

7. The intelligent visual rating method for fabric appearance smoothness according to claim 1, characterized in that: The step S6 comprises: (1) 3D point cloud generation: Using the depth value of each pixel in the depth image and combining it with the camera parameters, the pixel is converted from the 2D image space to the 3D world space to generate a 3D point cloud. (2) Point cloud processing: Smooth the generated 3D point cloud to eliminate noise and burrs; use statistical filtering and radius filtering to remove isolated points and noise points in the point cloud; (3) Triangular mesh generation: Using triangular mesh generation algorithms such as Poisson surface reconstruction, the processed point cloud is triangulated to generate a continuous triangular mesh model; (4) Post-processing: Optimize and adjust the initially generated triangular mesh model, map the texture information in the original fabric image to the triangular mesh model to enhance the realism of the model; optimize the edges and vertices of the triangular mesh to improve the fineness and smoothness of the model; (5) Visual analysis and display: Use 3D visualization software to visualize and analyze the generated triangular mesh model. Through the intuitive 3D model, the microscopic morphology and undulations of the fabric can be directly observed.

8. The intelligent visual rating method for fabric appearance smoothness according to claim 1, characterized in that: The step S8 comprises: (1) Data input and preprocessing: Input the fabric point cloud data of the new sample into the trained triangular mesh model; the input data should be in the same format as the training data, including 3D geometric features and preprocessed feature values; (2) Grade prediction: The triangular mesh model calculates the flatness grade of the fabric based on the input point cloud data through a deep learning algorithm and outputs a rating result from 1 to 5. During the prediction process, the model uses the relationship between the features and labels learned in the previous training to derive the corresponding flatness grade. (3) Result comparison and verification: The flatness grade predicted by the triangular mesh model is compared with the sample flatness grade obtained through subjective evaluation. At the same time, the real flatness data is used as a reference to verify whether the prediction of the triangular mesh model is consistent with the actual situation. (4) Accuracy optimization: In the feature extraction stage, high-resolution point cloud data and refined geometric feature analysis are used to capture the subtle height changes and wrinkle features on the fabric surface. The support vector machine algorithm is used for flatness rating, and during the rating, the cross-validation technology is used to verify the triangular mesh model multiple times. (5) Result output: The triangular mesh model will output the flatness rating of each sample and its corresponding geometric feature data.

9. An intelligent visual rating device for fabric appearance smoothness, characterized in that: It includes a three-dimensional frame, a binocular camera and an LED light source. The bottom of the three-dimensional frame is a fabric stage, the top of the three-dimensional frame is a top plate, the fabric stage is provided with a cloth stage, and the cloth stage is detachable; the LED light source is arranged on both sides of the binocular camera, the fabric stage is connected with a bottom stretching and extending component, the bottom stretching and extending component is used to extend the fabric stage, the bottom stretching and extending component is equipped with a homogeneous filling plate, and the homogeneous filling plate is filled in the bottom stretching and extending component after expansion; the top plate is connected with a top stretching and extending component, the top stretching and extending component is used to extend the top plate, the top stretching and extending component is equipped with a lightweight filling plate, and the lightweight filling plate is filled in the top stretching and extending component after expansion.

10. The intelligent visual rating device for fabric appearance smoothness according to claim 9, characterized in that: The top plate is provided with a built-in magnetic ring belt, and the bottom stretching and expansion component is provided with an external magnetic belt. The built-in magnetic ring belt and the external magnetic belt are used to fix the magnetic flexible curtain, and the magnetic flexible curtain is opaque.

Images