A visual inspection method based on fabric physical structure simulation

Abstract

This invention provides a visual inspection method based on fabric physical structure simulation. This method includes: a) Fabric physical modeling: establishing a physical structure model of the fabric, where represents the weave structure features, represents the material properties of the warp and weft yarns, and represents the arrangement of the warp and weft yarns; b) Automatic light source calibration: using sensors to detect the angle and intensity of the light source in real time, and dynamically adjusting the illumination angle and brightness of the light source through a calibration model to eliminate the influence of light source changes on detection. This visual inspection method based on fabric physical structure simulation innovatively solves the problem of the influence of changes in light angle and intensity on detection accuracy in existing technologies. Through multi-angle light source control and illumination compensation algorithms, using sensors to detect and calibrate the angle and intensity of the light source in real time ensures that the fabric surface exhibits consistent detection results under different lighting conditions.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of visual inspection, specifically a visual inspection method based on fabric physical structure simulation. BACKGROUND

[0002] The visual inspection method based on fabric physical structure simulation is mainly used for detecting fabric defects during the weaving process. By simulating the texture structure of the fabric, a simulated fabric or fabric blank is generated. Under the light irradiation, the fabric forms a base map with a 3-4 degree color difference as a contrast data source. The detection system compares this base map with the online generated fabric image to identify possible defects. Such a method can automatically monitor the quality during the fabric production process, thereby improving the production efficiency.

[0003] However, this detection method also has some defects. The angle and intensity of the light directly affect the accuracy of the detection, and if the lighting conditions change, it may lead to color difference judgment errors, thereby affecting the detection effect. The simulation effect of the system on the fabric texture may not perfectly cover the real situation, especially for complex fabric structures, the detection effect is weak. SUMMARY

[0004] In view of the defects of the prior art, the present application provides a visual inspection method based on fabric physical structure simulation, which solves the problem that the angle and intensity of the light directly affect the accuracy of the detection, and the simulation effect of the system on the fabric texture may not perfectly cover the real situation, especially for complex fabric structures, the detection effect is weak.

[0005] To achieve the above purpose, the present application realizes the following technical scheme: a visual inspection method based on fabric physical structure simulation, specifically comprising the following steps:

[0006] a. Fabric physical modeling: establish a physical structure model M=f(T,E,W) of the fabric, where T is the weaving structure characteristics, E is the warp and weft yarn material properties, and W is the arrangement of warp and weft yarns;

[0007] b. Light source automatic calibration: use sensors to detect the angle α and intensity I of the light source in real time, and dynamically adjust the irradiation angle and brightness of the light source through the calibration model C(α,I) to eliminate the influence of light source changes on detection;

[0008] c. Color difference detection area division: according to the fabric structure model M, divide the color difference detection area, and generate a color difference simulation image S r , and locally compare with the online detection image S t to improve the detection accuracy;

[0009] d. Image Acquisition and Preprocessing: A multi-angle, multi-light source camera system is used to acquire images of the fabric. Noise removal and contrast adjustment are performed using the preprocessing function P(·) to obtain the preprocessed image I. p =P(I raw );

[0010] e. Texture feature extraction: Based on the physical structure model of the fabric, extract texture features F = g(M), including the arrangement of warp and weft yarns, texture density and direction data, where g(·) is the texture feature extraction function;

[0011] f. Multi-angle light source control: Based on the texture features F and the light source calibration data C(α,I), the multi-angle light source is dynamically adjusted to ensure that the fabric can present clear texture features under different angles and lighting conditions;

[0012] g. Illumination compensation algorithm: For preprocessed image I p Illumination compensation is performed, the illumination difference compensation amount ΔL is calculated, and the compensated detection image I is obtained. c =I p -ΔL, to reduce the detection impact caused by uneven illumination;

[0013] h. Color difference detection and comparison: Based on the color difference detection algorithm, calculate the simulated image S r and compensated image I c Color difference ΔC = | S r -I c| Identify areas of color difference abnormality;

[0014] i. Defect Location: Based on the color difference detection result ΔC, determine the specific location P of the defect. w (x,y,z) and labeled in the fabric physical model;

[0015] j. Defect Feature Analysis: Perform feature analysis on the detected defects and calculate the defect type probability P(d). i |F), distinguish common defects such as broken yarn, missing stitches and misaligned stitches, so as to classify and archive them;

[0016] k. Data storage and feedback: The detection data and analysis results are stored in the system database, and the feedback data is sent back to the control system to adjust the production process parameters in order to improve production efficiency;

[0017] I. System Adaptive Control: Based on accumulated detection data, the system adaptively adjusts the light source and detection parameters through algorithms, automatically optimizes the detection process, and improves detection accuracy.

[0018] Preferably, the fabric physical modeling further adopts a multi-level refinement model containing micro and macro features to better approximate the real fabric structure, and the light source automatic calibration module includes light source intensity sensors and angle sensors to ensure stable light sources in different environments.

[0019] Preferably, the color difference detection area division is based on an automatic generation algorithm of fabric weaving structure, enabling each area to respond to complex textures, and the image acquisition device adopts a multi-band imaging mode to collect multiple spectral images of the same detection area, improving detection accuracy.

[0020] Preferably, the texture feature extraction adopts a machine learning algorithm to automatically identify and classify different texture features of the fabric, and the multi-angle light source control system automatically adjusts the light source angle and brightness according to the color difference changes of the fabric to adapt to the detection needs of different fabrics.

[0021] Preferably, the light compensation algorithm further includes a region-based dynamic compensation algorithm to adapt to uneven image lighting, and the color difference detection and contrast algorithm is based on a deep learning model to identify small color difference changes to improve detection sensitivity.

[0022] Preferably, the defect positioning is based on three-dimensional coordinate labeling for accurate recording of defect positions, and the defect feature analysis module includes a defect sample database for storing detected defect samples to improve system recognition ability.

[0023] Preferably, the data storage module is integrated with a real-time feedback system to directly feed the defect detection data to the production equipment for real-time adjustment, and the production process adjustment module adjusts the production parameters adaptively through statistical analysis of the detection data.

[0024] Preferably, the color difference detection algorithm has an adaptive threshold adjustment function to meet the detection needs of different fabric materials, and the detection system is suitable for various fabric weaving methods and can automatically adapt to fabrics with different warp and weft densities, expanding the application range of the system.

[0025] The present application provides a visual detection method based on fabric physical structure simulation, which has the following advantages:

[0026] The visual detection method based on fabric physical structure simulation innovatively solves the influence of light angle and intensity change on detection accuracy in the prior art. Through multi-angle light source control and illumination compensation algorithm, the angle and intensity of the light source are detected and calibrated in real time by the sensor, ensuring that the fabric surface can present consistent detection effect under different illumination conditions. The scheme builds a detailed model of the fabric physical structure, covers the fabric weaving structure and the warp and weft yarn material characteristics, and realizes texture feature extraction combined with machine learning algorithm, which can more effectively simulate the real texture characteristics of the fabric, especially improves the detection effect of complex fabric structure.

[0027] The advantage of the present application lies in its automatic calibration and adaptive detection method, which can automatically adjust the light source and color difference detection parameters to ensure the detection accuracy in complex production environment. In addition, the fabric is physically modeled by a multi-level detailed model, so that the detection system adapts to different fabric materials and weaving methods, enhancing the applicability of the system. At the same time, the technical scheme also has the function of accurate defect positioning based on three-dimensional coordinates, and the defect data is fed back to the production equipment in real time for automatic adjustment, thereby effectively improving the production efficiency and product quality. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 The flowchart of the present application is shown. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0030] As Figure 1 shown, the embodiment of the present application provides a visual detection method based on fabric physical structure simulation, which comprises: a. fabric physical modeling: establishing a physical structure model M=f(T,E,W) of the fabric, wherein T is the weaving structure feature, E is the warp and weft yarn material characteristics, and W is the arrangement of warp and weft yarns. The fabric physical modeling further adopts a multi-level detailed model, including micro and macro features, so as to be closer to the real fabric structure. The light source automatic calibration module includes a light source intensity sensor and an angle sensor to ensure stable light source under different environments.

[0031] b. Light source automatic calibration: real-time detection of the angle a and intensity I of the light source by the sensor, dynamic adjustment of the illumination angle and brightness of the light source through the calibration model C(a,I) to eliminate the influence of light source change on detection.

[0032] c. Color difference detection area division: Based on the fabric structure model M, the color difference detection area is divided, and a color difference simulation image S is generated. r , with online detection image S t Local comparison is performed to improve detection accuracy. The color difference detection area division is based on an automatic generation algorithm of the fabric weave structure, which enables each area to respond to complex textures. The image acquisition device adopts a multi-band imaging mode to acquire multiple spectral images of the same detection area, thereby improving detection accuracy.

[0033] d. Image Acquisition and Preprocessing: A multi-angle, multi-light source camera system is used to acquire images of the fabric. Noise removal and contrast adjustment are performed using the preprocessing function P(·) to obtain the preprocessed image I. p =P(I raw ).

[0034] e. Texture feature extraction: Based on the physical structure model of the fabric, the texture feature F = g(M) is extracted, including the arrangement of warp and weft yarns, texture density and direction data. g(·) is the texture feature extraction function. The texture feature extraction adopts machine learning algorithm to automatically identify and classify different texture features of the fabric. The multi-angle light source control system automatically adjusts the light source angle and brightness according to the color difference change of the fabric to adapt to the detection needs of different fabrics.

[0035] f. Multi-angle light source control: Based on the texture features F and the light source calibration data C(α,I), the multi-angle light source is dynamically adjusted to ensure that the fabric can present clear texture features under different angles and lighting conditions;

[0036] g. Illumination compensation algorithm: For preprocessed image I p Illumination compensation is performed, the illumination difference compensation amount ΔL is calculated, and the compensated detection image I is obtained. c =I p -ΔL is used to reduce the detection impact of uneven illumination. The illumination compensation algorithm further includes a region-based dynamic compensation algorithm to adapt to uneven image illumination. The color difference detection and comparison algorithm is based on a deep learning model to identify minute color difference changes in order to improve detection sensitivity.

[0037] h. Color difference detection and comparison: Based on the color difference detection algorithm, calculate the simulated image S r and compensated image I c Color difference ΔC = | S r -I c | Identify areas with abnormal color differences.

[0038] i. Defect Location: Based on the color difference detection result ΔC, determine the specific location P of the defect. w(x, y, z) and mark in the fabric physical model, the defect positioning is based on three-dimensional coordinate marking, which is used for accurately recording the defect position, the defect feature analysis module includes a defect sample database for storing detected defect samples, and the system recognition ability is improved.

[0039] j. Defect feature analysis: feature analysis is performed on the detected defects, and the defect type probability P is calculated ( d i ∣F ) , common defects such as broken yarn, missing knitting and complex are distinguished for classification and archiving.

[0040] k. Data storage and feedback: store the detection data and analysis results in the system database, and return the feedback data to the control system to adjust the production process parameters to improve the production efficiency, the data storage module and the real-time feedback system are integrated, and the defect detection data can be directly fed back to the production equipment for real-time adjustment, and the production process adjustment module adjusts the production parameters through statistical analysis of the detection data.

[0041] I. System adaptive control: based on the cumulative detection data, the light source and detection parameters are automatically adjusted through algorithm, the detection process is automatically optimized, the detection accuracy is improved, the color difference detection algorithm has the function of adaptive threshold adjustment to meet the detection needs of different fabric materials, the detection system is suitable for various fabric knitting methods, and can automatically adapt to fabrics with different warp and weft densities, and the application range of the system is expanded.

[0042] Although the embodiments of the present application have been shown and described, it can be understood by those skilled in the art that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the present application, the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A visual inspection method based on fabric physical structure simulation, characterized in that, Specifically comprising the following steps: a. Fabric physical modeling: Establish a physical structure model M = f(T, E, W) of the fabric, where T is the weaving structure feature, E is the warp and weft material property, and W is the arrangement of warp and weft; b. Light source automatic calibration: Real-time detection of the angle α and intensity I of the light source using sensors, dynamic adjustment of the angle and intensity of the light source through the calibration model C(α, I) to eliminate the influence of light source changes on detection; c. Color difference detection region division: according to the physical structure model M of the fabric, the region for color difference detection is divided, and a simulated image S for color difference detection is generated r , and the online detection image S t are locally compared; d. Image acquisition and preprocessing: Using a multi-angle, multi-light source camera system to acquire images of the fabric, removing noise and adjusting contrast through a preprocessing function to obtain preprocessed images; e. Texture feature extraction: Based on the physical structure model M of the fabric, extract texture features, including the arrangement of warp and weft, texture density and direction data; f. Multi-angle light source control: According to the texture features and the light source calibration model C(α, I), dynamically adjust the multi-angle light source to ensure that the fabric can present clear texture features under different angles and lighting; g. Light compensation algorithm: light compensation is performed on the preprocessed image to obtain a compensated detection image I c to reduce the detection influence caused by uneven illumination; h. Color difference detection and contrast: based on the color difference detection algorithm, calculate the color difference ΔC= S r and the compensated detection image I c of the simulation image S | S r -I c| , identify the color difference abnormal area; i. Defect location: according to the color difference AC, the specific location P of the defect is determined w (x, y, z) and marked in the physical structure model of the fabric; j. Defect feature analysis: Feature analysis of detected defects, calculating the probability of defect type, including broken yarn, missing weaving and complex.

2. The visual inspection method based on fabric physical structure simulation according to claim 1, characterized in that: The fabric physical modeling further adopts a multi-level refinement model, containing micro and macro features, and the light source automatic calibration sensor includes a light source intensity sensor and an angle sensor.

3. The visual inspection method based on fabric physical structure simulation according to claim 1, characterized in that: The color difference detection region division is based on an automatic generation algorithm based on the fabric weaving structure, so that each region can respond to complex textures, and the image acquisition adopts a multi-band imaging mode to acquire multiple spectral images of the same detection area.

4. The visual inspection method based on fabric physical structure simulation according to claim 1, characterized in that: The texture feature extraction adopts a machine learning algorithm to automatically identify and classify different texture features of the fabric, and the multi-angle light source control automatically adjusts the light source angle and brightness according to the color difference changes of the fabric.

5. The visual inspection method based on fabric physical structure simulation according to claim 1, characterized in that: The illumination compensation algorithm further includes a region-based dynamic compensation algorithm to adapt to uneven image lighting, and the color difference detection and contrast are based on a deep learning model to identify small color difference changes.

6. The visual inspection method based on fabric physical structure simulation according to claim 1, characterized in that: The defect positioning is based on three-dimensional coordinate labeling for accurate recording of defect location, and the defect feature analysis includes a defect sample database for storing detected defect samples.

7. The visual inspection method based on fabric physical structure simulation according to claim 1, characterized in that: The color difference detection algorithm has an adaptive threshold adjustment function to meet the detection needs of different fabric materials, and the detection method is suitable for various fabric weaving methods and can automatically adapt to fabrics with different warp and weft densities.

Images