Ribbon quality inspection system and method based on industrial camera inspection

Abstract

The present invention provides a webbing quality detection system and method based on industrial camera detection, which relate to the technical field of data processing. The system comprises: a speed tracking module; a multi-angle acquisition module; an image fusion module; an optical correction module; a preprocessing module; a preliminary defect detection module; a fine defect analysis module; an adaptive filtering module; a dynamic coordinate mapping module, which is used to map the physical position of the defect and obtain actual coordinate data; a data storage and analysis module, which is used to store and analyze historical actual coordinate data and related detection data, and generate a quality report and a trend analysis report; an alarm and control module, which is used to automatically send out an alarm signal when the number of defects or the trend change rate in the trend analysis report reaches a preset threshold, and adjust production parameters or suspend production. The present invention improves the accuracy and efficiency of webbing quality detection.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a webbing quality detection system and method based on industrial camera detection. Background Art

[0002] Existing webbing quality inspection technology mainly relies on manual visual inspection, where operators visually observe the flaws and defects on the webbing surface, such as breakage, burrs, color difference, etc. Although this method can detect defects in the webbing to a certain extent, it relies on the subjective judgment of the operator and is easily affected by factors such as fatigue and distraction, resulting in unstable and inconsistent inspection results.

[0003] In order to solve the shortcomings of manual inspection, automated visual inspection systems based on industrial cameras have begun to be introduced in the existing technology. This type of system uses industrial cameras to collect webbing images, and combines image processing algorithms to automatically identify and analyze possible defects on the webbing surface, greatly improving the speed and consistency of inspection. However, the existing industrial camera-based inspection systems still have some technical limitations. For example, in a fast-moving production line, the high-speed movement of the webbing often causes blurring or distortion during image acquisition. Especially in the case of complex textures or reflections on the webbing surface, traditional image processing algorithms are difficult to effectively distinguish between normal textures and defects, resulting in the accuracy of the inspection results being affected. Summary of the invention

[0004] The purpose of the present invention is to provide a webbing quality detection system and method based on industrial camera detection, aiming to solve the problems mentioned in the background technology.

[0005] In order to solve the above technical problems, the technical solution of the present invention is as follows:

[0006] In the first aspect, a webbing quality inspection system based on industrial camera inspection, the system comprises:

[0007] Speed ​​tracking module, used to monitor the moving speed of the webbing in real time and generate synchronization signals;

[0008] The multi-angle acquisition module is used to capture multi-angle images of the same position from different angles through multiple groups of industrial cameras according to synchronization signals to obtain initial image data;

[0009] An image fusion module is used to spatially align and fuse the initial image data to generate a high-resolution image;

[0010] An optical correction module is used to optically compensate the surface texture of the ribbon in the high-resolution image according to the angle difference of the collected light to obtain a corrected image;

[0011] A preprocessing module is used to perform denoising and enhancement processing on the corrected image to obtain an analysis image;

[0012] A preliminary defect detection module is used to perform preliminary defect detection on the analysis image according to a preset comparison image, and to preliminarily mark significant areas that deviate from the normal structure to obtain a rough screening area image;

[0013] The fine defect analysis module is used to fit the edge contour of the coarse screening area to obtain the defect recognition area image, analyze the internal texture and shape characteristics of the defect recognition area image, identify potential minor defects, and generate a fine defect image;

[0014] Adaptive filtering module, used to filter fine defect images according to the material type and surface characteristics of the ribbon to obtain defect coordinate data;

[0015] The dynamic coordinate mapping module is used to accurately map the physical position of the defect by dynamically associating the defect coordinate data with the motion trajectory of the ribbon to obtain the actual coordinate data;

[0016] Data storage and analysis module, used to store and analyze historical actual coordinate data and related test data, and generate quality reports and trend analysis reports;

[0017] The alarm and control module is used to automatically send out an alarm signal and adjust production parameters or suspend production when the number of defects or trend change rate in the trend analysis report reaches a preset threshold.

[0018] Preferably, the image fusion module comprises:

[0019] A transformation matrix generation unit, used for obtaining a position vector and a rotation matrix of each industrial camera and calculating a space transformation matrix of each camera according to the position vector and the rotation matrix of each industrial camera;

[0020] A spatial alignment unit, used for performing coordinate transformation on the initial image data according to the spatial transformation matrix to obtain an aligned image;

[0021] The weighted fusion unit is used to fuse the aligned images according to the shooting angle and clarity of each industrial camera to obtain a high-resolution image. The calculation formula of the pixel value of the high-resolution image is:

[0022] ,

[0023] in, For high-resolution images at coordinates The pixel value at is the number of industrial cameras for image acquisition, For Camera The weight of image fusion, where ,in The camera shooting angle, is the camera’s resolution factor, is a normalization term used to ensure that the sum of the weights of all cameras is 1. For the Camera in The aligned image pixel value at position, is the fusion coefficient, , is the edge enhancement coefficient, which is used to balance the pixel value and edge information. For the The camera captures the image in Laplacian of position;

[0024] in, , For the The standard deviation of the images taken by the cameras, is the weight coefficient, which is used to control the influence of camera clarity on edge clarity. For the The edge strength measure of the camera image, ,in and is the number of rows and columns of the image, and Respectively represent Camera images in and Direction gradient.

[0025] Preferably, the optical correction module comprises:

[0026] An angle compensation unit is used to obtain the shooting angle data of each industrial camera, calculate the light deviation value at different angles according to the shooting angle data, and generate angle compensation parameters;

[0027] The light correction unit is used to adjust the brightness value and color information of each pixel in the high-resolution image according to the angle compensation parameter to obtain a corrected image; wherein the calculation formula of the pixel value of the corrected image is:

[0028] ,

[0029] in, To correct the image in coordinates The pixel value at is the brightness compensation coefficient, is the incident angle of the light, is the color correction coefficient, is the angle influence coefficient, For shooting angle, is the edge correction coefficient, and The images are and The second partial derivative in the direction.

[0030] Preferably, the preprocessing module comprises:

[0031] A noise suppression unit is used to detect and remove image noise in the modified image, reduce image noise interference through an edge-preserving noise reduction algorithm, and obtain a noise-suppressed image;

[0032] A regional contrast enhancement unit is used to analyze the brightness difference and contrast of each pixel in the noise suppression image, improve the contrast of the subtle texture on the surface of the ribbon, and obtain a first enhanced image;

[0033] An edge refinement unit, configured to refine edge features of the ribbon texture using an edge enhancement algorithm according to the first enhanced image, so as to generate a second enhanced image;

[0034] The multi-scale detail processing unit is used to decompose detail information of the second enhanced image at multiple scales, perform contrast enhancement processing on details of each scale respectively to obtain enhanced information, and then fuse the enhanced information of different scales to generate an analysis image.

[0035] Preferably, the preliminary defect detection module comprises:

[0036] The multi-level structure comparison unit is used to compare features between the analysis image and the preset comparison image, calculate the similarity, and identify the first area that deviates from the normal structure; the feature similarity calculation formula is as follows:

[0037] ,

[0038] in, For the coordinates The feature similarity at is the number of multi-scale layers, For the The weight of the layer scale, ,in is the weight coefficient, is the normalization term, and In the Analyzing images at the layer scale Compare image with preset The mean of and In the Analyzing images at the layer scale Compare image with preset The variance of To analyze the image Compare image with preset In the Covariance at the layer scale, and is a smoothing constant used to prevent the denominator from being zero;

[0039] The defect area marking unit is used to screen and mark the first area through a preset similarity threshold to generate a rough screening area image.

[0040] Preferably, the fine defect analysis module includes:

[0041] A contour fitting unit is used to accurately fit the edge of the rough screening area image to generate a fine edge image;

[0042] The internal structure analysis unit is used to analyze the texture and shape characteristics of the edge fine image, identify the area containing tiny defects, and form a fine defect image.

[0043] Preferably, the adaptive filtering module comprises:

[0044] A material property identification unit, used to automatically select a preset corresponding defect filtering parameter set according to the material type of the webbing;

[0045] The defect screening unit is used to screen each defect area in the fine defect image according to the defect filtering parameter set, and convert it into defect coordinate data of the defect area.

[0046] Preferably, the dynamic coordinate mapping module comprises:

[0047] The motion trajectory monitoring unit is used to monitor the motion trajectory of the webbing in the production process in real time and generate the position data of the webbing in the production process;

[0048] The dynamic position mapping unit is used to associate the defect coordinate data with the position data, determine the dynamic position of the defect in the physical space, and obtain the actual coordinate data.

[0049] Preferably, the data storage and analysis module includes:

[0050] A historical data storage unit, used to record and store actual coordinate data and related test data to form a historical data set of webbing quality;

[0051] The data analysis unit performs statistical analysis on historical data sets, identifies the changing trends and quality issues of webbing quality, and generates quality reports and trend analysis reports.

[0052] In the second aspect, a method for detecting the quality of a webbing based on industrial camera detection comprises the following steps:

[0053] Monitor the moving speed of the webbing in real time and generate synchronization signals;

[0054] Using multiple groups of industrial cameras to capture multi-angle images of the same position from different angles according to synchronization signals, initial image data is obtained;

[0055] Spatially align and fuse the initial image data to generate a high-resolution image;

[0056] Optically compensate the surface texture of the ribbon in the high-resolution image according to the angle difference of the collected light to obtain a corrected image;

[0057] De-noising and enhancing the corrected image to obtain an analysis image;

[0058] Perform preliminary defect detection on the analysis image according to the preset comparison image, preliminarily mark the significant areas that deviate from the normal structure, and obtain a rough screening area image;

[0059] Perform edge contour fitting on the coarse screening area to obtain the defect recognition area image, and analyze the internal texture and shape characteristics of the defect recognition area image to identify potential minor defects and generate a fine defect image;

[0060] According to the material type and surface characteristics of the ribbon, the fine defect image is filtered to obtain the defect coordinate data;

[0061] By dynamically associating the defect coordinate data with the motion trajectory of the ribbon, the physical position of the defect is accurately mapped to obtain the actual coordinate data;

[0062] Store and analyze historical actual coordinate data and related test data, and generate quality reports and trend analysis reports;

[0063] When the number of defects or the trend change rate in the trend analysis report reaches the preset threshold, an alarm signal is automatically issued, and production parameters are adjusted or production is suspended.

[0064] The above solution of the present invention includes at least the following beneficial effects:

[0065] The system first monitors the moving speed of the webbing in real time through the speed tracking module and generates a synchronization signal to control the synchronous operation of multiple industrial cameras of the multi-angle acquisition module. Under the guidance of the synchronization signal, the multi-angle acquisition module can capture images of the webbing surface from multiple angles to ensure the integrity of the image capture and the comprehensiveness of the detailed information. This setting solves the problem of information loss under traditional single-angle acquisition, ensuring that the system can still capture clear image data even when the webbing moves at high speed, effectively reducing the possibility of image distortion and misalignment.

[0066] The image fusion module performs spatial alignment and fusion on the initial image data collected from multiple angles to generate a high-resolution image. Spatial alignment ensures that images collected from different angles overlap in the same coordinate system, avoiding image misalignment caused by differences in shooting angles in traditional systems. The high-resolution image generated by the fusion process greatly enhances the image details, allowing subsequent inspection modules to perform defect analysis on a clearer data source, improving the accuracy of inspection. The optical correction module further optimizes the optical consistency of the image, eliminates the brightness and color differences caused by different shooting angles through light angle compensation, obtains a more uniform corrected image, and avoids misjudgment caused by differences in lighting conditions.

[0067] In the preprocessing module, the system performs denoising and enhancement on the corrected image. The generated analysis image has high contrast and clarity, providing a reliable image basis for the defect detection module. The preliminary defect detection module uses the analysis image to compare the structure with the preset comparison image, preliminarily marks the significant areas that deviate from the normal structure, and generates a rough screening area image. This process ensures the accuracy of the preliminary screening and lays the foundation for subsequent fine analysis. On the basis of coarse screening, the fine defect analysis module fits the edges of the defective area and further analyzes the texture and shape features to identify potential minor defects and generate a fine defect image. Through this in-depth analysis, the system can accurately locate and classify various defects on the surface of the ribbon, greatly improving the accuracy and meticulousness of detection.

[0068] The adaptive filtering module performs targeted filtering on fine defect images according to the material and surface characteristics of the webbing, retaining only the coordinate data that meets the defect characteristics and eliminating the interference caused by material noise. The dynamic coordinate mapping module dynamically associates the filtered defect coordinate data with the motion trajectory of the webbing, accurately maps the position of the defect in the physical space, and ensures the traceability and positioning accuracy of the defect data in practical applications. The data storage and analysis module stores and analyzes the actual coordinate data and detection data to generate a historical trend report on the webbing quality. This big data analysis method can predict quality problems in advance and provide valuable decision-making basis for production management. When the system detects that the number of defects or the trend change rate reaches the threshold, the alarm and control module automatically triggers an alarm signal, and even adjusts production parameters or suspends production as needed to control the spread of defects in a timely manner.

[0069] The multi-module setting of this system enables it to have higher accuracy and automation level in detecting the quality of webbing. Compared with the existing technology, the system realizes more efficient defect identification and control in the whole process of collection, processing, analysis and feedback, which provides important guarantee for the efficient production and quality management of webbing. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 It is an architecture diagram of a webbing quality inspection system based on industrial camera inspection provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0071] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0072] like Figure 1 As shown, an embodiment of the present invention proposes a ribbon quality detection system based on industrial camera detection, and the system includes:

[0073] Speed ​​tracking module, used to monitor the moving speed of the webbing in real time and generate synchronization signals;

[0074] The multi-angle acquisition module is used to capture multi-angle images of the same position from different angles through multiple groups of industrial cameras according to synchronization signals to obtain initial image data;

[0075] An image fusion module is used to spatially align and fuse the initial image data to generate a high-resolution image;

[0076] An optical correction module is used to optically compensate the surface texture of the ribbon in the high-resolution image according to the angle difference of the collected light to obtain a corrected image;

[0077] A preprocessing module is used to perform denoising and enhancement processing on the corrected image to obtain an analysis image;

[0078] A preliminary defect detection module is used to perform preliminary defect detection on the analysis image according to a preset comparison image, and to preliminarily mark significant areas that deviate from the normal structure to obtain a rough screening area image;

[0079] The fine defect analysis module is used to fit the edge contour of the coarse screening area to obtain the defect recognition area image, analyze the internal texture and shape characteristics of the defect recognition area image, identify potential minor defects, and generate a fine defect image;

[0080] Adaptive filtering module, used to filter fine defect images according to the material type and surface characteristics of the ribbon to obtain defect coordinate data;

[0081] The dynamic coordinate mapping module is used to accurately map the physical position of the defect by dynamically associating the defect coordinate data with the motion trajectory of the ribbon to obtain the actual coordinate data;

[0082] Data storage and analysis module, used to store and analyze historical actual coordinate data and related test data, and generate quality reports and trend analysis reports;

[0083] The alarm and control module is used to automatically send out an alarm signal and adjust production parameters or suspend production when the number of defects or trend change rate in the trend analysis report reaches a preset threshold.

[0084] In the embodiment of the present invention, the system first monitors the moving speed of the webbing in real time through the speed tracking module, and generates a synchronization signal to control the synchronous operation of multiple industrial cameras of the multi-angle acquisition module. Under the guidance of the synchronization signal, the multi-angle acquisition module can capture images of the webbing surface from multiple angles to ensure the integrity of the image capture and the comprehensiveness of the detailed information. This setting solves the problem of information loss under traditional single-angle acquisition, ensuring that the system can still capture clear image data even when the webbing moves at high speed, effectively reducing the possibility of image distortion and misalignment.

[0085] The image fusion module performs spatial alignment and fusion on the initial image data collected from multiple angles to generate a high-resolution image. Spatial alignment ensures that images collected from different angles overlap in the same coordinate system, avoiding image misalignment caused by differences in shooting angles in traditional systems. The high-resolution image generated by the fusion process greatly enhances the image details, allowing subsequent inspection modules to perform defect analysis on a clearer data source, improving the accuracy of inspection. The optical correction module further optimizes the optical consistency of the image, eliminates the brightness and color differences caused by different shooting angles through light angle compensation, obtains a more uniform corrected image, and avoids misjudgment caused by differences in lighting conditions.

[0086] In the preprocessing module, the system performs denoising and enhancement on the corrected image. The generated analysis image has high contrast and clarity, providing a reliable image basis for the defect detection module. The preliminary defect detection module uses the analysis image to compare the structure with the preset comparison image, preliminarily marks the significant areas that deviate from the normal structure, and generates a rough screening area image. This process ensures the accuracy of the preliminary screening and lays the foundation for subsequent fine analysis. On the basis of coarse screening, the fine defect analysis module fits the edges of the defective area and further analyzes the texture and shape features to identify potential minor defects and generate a fine defect image. Through this in-depth analysis, the system can accurately locate and classify various defects on the surface of the ribbon, greatly improving the accuracy and meticulousness of detection.

[0087] The adaptive filtering module performs targeted filtering on fine defect images according to the material and surface characteristics of the webbing, retaining only the coordinate data that meets the defect characteristics and eliminating the interference caused by material noise. The dynamic coordinate mapping module dynamically associates the filtered defect coordinate data with the motion trajectory of the webbing, accurately maps the position of the defect in the physical space, and ensures the traceability and positioning accuracy of the defect data in practical applications. The data storage and analysis module stores and analyzes the actual coordinate data and detection data to generate a historical trend report on the webbing quality. This big data analysis method can predict quality problems in advance and provide valuable decision-making basis for production management. When the system detects that the number of defects or the trend change rate reaches the threshold, the alarm and control module automatically triggers an alarm signal, and even adjusts production parameters or suspends production as needed to control the spread of defects in a timely manner.

[0088] The multi-module setting of this system enables it to have higher accuracy and automation level in detecting the quality of webbing. Compared with the existing technology, the system realizes more efficient defect identification and control in the whole process of collection, processing, analysis and feedback, which provides important guarantee for the efficient production and quality management of webbing.

[0089] In a preferred embodiment of the present invention, the image fusion module includes:

[0090] A transformation matrix generation unit, used for obtaining a position vector and a rotation matrix of each industrial camera and calculating a space transformation matrix of each camera according to the position vector and the rotation matrix of each industrial camera;

[0091] A spatial alignment unit, used for performing coordinate transformation on the initial image data according to the spatial transformation matrix to obtain an aligned image;

[0092] The weighted fusion unit is used to fuse the aligned images according to the shooting angle and clarity of each industrial camera to obtain a high-resolution image. The calculation formula of the pixel value of the high-resolution image is:

[0093] ,

[0094] in, For high-resolution images at coordinates The pixel value at is the number of industrial cameras for image acquisition, For Camera The weight of image fusion, where ,in The camera shooting angle, is the camera’s resolution factor, is a normalization term used to ensure that the sum of the weights of all cameras is 1. For the Camera in The aligned image pixel value at position, is the fusion coefficient, , is the edge enhancement coefficient, which is used to balance the pixel value and edge information. For the The camera captures the image in Laplacian of position;

[0095] in, , For the The standard deviation of the images taken by the cameras, is the weight coefficient, which is used to control the influence of camera clarity on edge clarity. For the The edge strength measure of the camera image, ,in and is the number of rows and columns of the image, and Respectively represent Camera images in and Direction gradient.

[0096] In an embodiment of the present invention, the image fusion module uses a transformation matrix generation unit, a spatial alignment unit, and a weighted fusion unit to perform high-quality integration of the initial image data. First, the transformation matrix generation unit generates a corresponding spatial transformation matrix by calculating the position vector and rotation matrix of each industrial camera, thereby converting the data collected from multiple angles into the same coordinate space. The spatial alignment unit uses the spatial transformation matrix to perform coordinate conversion on the initial image data of each camera to ensure that the data of all cameras can accurately overlap at the same position. This processing method effectively eliminates the deviation problem caused by multi-angle shooting. Through this spatial alignment method, the fused high-resolution image avoids distortion or dislocation, laying a solid foundation for subsequent image processing.

[0097] In the weighted fusion unit, the system performs weighted fusion processing on the aligned images according to the shooting angle and clarity of each industrial camera. Indicates The weight adjustment result of the image of the camera after brightness adjustment and edge enhancement is obtained by adjusting the parameters The data from different cameras can be balanced. This weighted fusion method can more effectively highlight the details of the clear image, while improving the coherence of the image at the edge, making the high-resolution image have higher quality under the integration of multi-angle information.

[0098] Specifically, the following is a method for obtaining each coefficient in this embodiment:

[0099] For edge enhancement factor , can be adjusted by experimental determination on a series of test images values ​​to ensure that the fused image achieves the expected effect in terms of edge sharpness. For example, by testing different The effect of the value on edge sharpening can be compared and optimized by objective image quality indicators such as peak signal-to-noise ratio (PSNR) or structural similarity (SSIM). In addition, edge detection algorithms in image processing, such as Sobel operator or Canny edge detection, can be used to measure edge strength and use the measurement results to dynamically adjust the For example, according to the edge strength, the appropriate value.

[0100] For the adjustment factor , you can manually set a fixed value based on the performance of different cameras and empirical data on shooting distances. value to ensure that the clarity factor is adjusted reasonably. It can also be gradually optimized through experiments with different clarity requirements. The value can be adjusted to find the best balance between edge clarity and overall image quality. In addition, the experimental data can be automatically adjusted through cross-validation. The image quality is evaluated under different values, and the one that can achieve the best fusion effect is finally selected. value.

[0101] Specifically, the transformation matrix generation unit is used to obtain the position vector and rotation matrix of each industrial camera, and calculate the spatial transformation matrix of each camera based on this. The specific steps are as follows: First, according to the installation position and direction of each camera, measure and record the three-dimensional position coordinates of the camera, that is, the position vector, and the rotation information of the camera, that is, the rotation matrix. These data are collected by calibration tools or software. Then, the position vector and the rotation matrix are combined, and a spatial transformation matrix is ​​constructed through homogeneous coordinate transformation, so that the image taken by each camera can be transformed into a unified three-dimensional space through this matrix. This spatial transformation matrix is ​​used for subsequent image alignment to ensure that data collected from multiple angles can be unified in the same coordinate system.

[0102] The spatial alignment unit uses the spatial transformation matrix calculated by the transformation matrix generation unit to perform coordinate transformation on the initial image data to obtain an aligned image. The spatial alignment process transforms the pixel coordinates in each image into a unified spatial coordinate system, thereby achieving spatial consistency of images from different camera perspectives. In actual operation, the original pixel coordinates are mapped to the alignment coordinates in the standard coordinate system by performing a matrix transformation on each pixel position. This alignment process can effectively eliminate the deviation caused by the difference in multi-camera angles, laying the foundation for subsequent image fusion.

[0103] In a preferred embodiment of the present invention, the optical correction module comprises:

[0104] An angle compensation unit is used to obtain the shooting angle data of each industrial camera, calculate the light deviation value at different angles according to the shooting angle data, and generate angle compensation parameters;

[0105] The light correction unit is used to adjust the brightness value and color information of each pixel in the high-resolution image according to the angle compensation parameter to obtain a corrected image; wherein the calculation formula of the pixel value of the corrected image is:

[0106] ,

[0107] in, To correct the image in coordinates The pixel value at is the brightness compensation coefficient, is the incident angle of the light, is the color correction coefficient, is the angle influence coefficient, For shooting angle, is the edge correction coefficient, and The images are and The second partial derivative in the direction.

[0108] In an embodiment of the present invention, the image fusion module performs spatial alignment and fusion of multi-angle images of the webbing based on the initial image data to generate a high-resolution image. The role of spatial alignment is to eliminate the coordinate deviation caused by different shooting angles, so that the data collected from multiple angles can be unified in the same spatial coordinate system and ensure the continuity between the images. The high-resolution image not only enhances the details of the webbing, but also enables the subsequent optical correction, preprocessing and defect detection modules to be processed on the basis of a higher quality data source. Compared with the single acquisition angle or asynchronous acquisition method in the prior art, this method effectively improves the ability to present details of the detection. The generation of the high-resolution image provides more abundant information for subsequent image processing and defect analysis, which is conducive to further improving the accuracy and stability of defect identification.

[0109] The optical correction module, in cooperation with the angle compensation unit and the light correction unit, effectively corrects the optical deviation of the high-resolution image. The angle compensation unit calculates the angle deviation value based on the shooting angle data of each camera, generates angle compensation parameters, and then compensates for the light differences at different angles. This ensures that the surface texture of the ribbon in the high-resolution image can maintain consistent optical performance at different shooting angles, thereby reducing the impact of brightness and color differences on the image. The light correction unit further adjusts the brightness and color information of each pixel of the image based on the angle compensation parameters to ensure the visual consistency of the high-resolution image and provide a unified data basis for subsequent analysis.

[0110] In the light correction formula, Used to combine brightness compensation coefficient and color correction coefficients , based on the incident angle of the light Non-linear compensation is used to correct brightness and color deviations. The angle influence coefficient is and edge correction factor The method ensures the optical consistency of high-resolution images under different lighting conditions, making the optical compensation effect more accurate.

[0111] Specifically, the following is a method for obtaining each coefficient in this embodiment:

[0112] For brightness compensation coefficient , can be determined by correcting for brightness changes under different ambient light conditions. For example, a larger You can also adjust the brightness compensation effect experimentally to test different The value is used to measure the performance of the image under various shooting conditions, so as to find an optimal value suitable for all scenes.

[0113] For color correction coefficients , can be set according to the actual color deviation value collected during camera calibration , ensuring that the colors in the image are as close to reality as possible. You can also use color calibration cards or specific standard color models, such as sRGB, to adjust the color in the calibration test. values ​​to ensure minimal color deviation under different lighting conditions.

[0114] For the angle influence coefficient , can be gradually adjusted in the actual environment To observe its effect on image edge clarity and brightness uniformity. You can also use optical simulation software to simulate the incident angle and brightness attenuation of light, and get the attenuation curve at different angles to determine the appropriate value.

[0115] For edge correction factor , image edge quality assessment indicators such as gradient amplitude or contrast can be used to set The value can be adjusted to balance the edge sharpening effect and overall image smoothness. It can also automatically adjust the image quality by training machine learning models in large amounts of image data. value to optimize the edge enhancement effect.

[0116] Specifically, the main function of the angle compensation unit is to obtain the shooting angle data of each industrial camera, and based on this data, calculate the light deviation value at different angles, and generate corresponding angle compensation parameters to ensure the optical consistency of high-resolution images at different shooting angles. First, the shooting angle of each camera is measured by the inclination sensor or calibration system installed on the camera, and the angle information is recorded. Then, according to the angle between the different shooting angles and the normal of the ribbon surface, the change in the incident angle of the light is calculated, and the brightness and color deviation caused by the angle is estimated. Finally, these deviation data are used to generate angle compensation parameters, which will be used in subsequent light correction to adjust the brightness and color information in the image. This can effectively reduce the optical deviation caused by differences in shooting angles and ensure the consistency of the image in visual effects.

[0117] In a preferred embodiment of the present invention, the preprocessing module comprises:

[0118] A noise suppression unit is used to detect and remove image noise in the modified image, reduce image noise interference through an edge-preserving noise reduction algorithm, and obtain a noise-suppressed image;

[0119] A regional contrast enhancement unit is used to analyze the brightness difference and contrast of each pixel in the noise suppression image, improve the contrast of the subtle texture on the surface of the ribbon, and obtain a first enhanced image;

[0120] An edge refinement unit, configured to refine edge features of the ribbon texture using an edge enhancement algorithm according to the first enhanced image, so as to generate a second enhanced image;

[0121] The multi-scale detail processing unit is used to decompose detail information of the second enhanced image at multiple scales, perform contrast enhancement processing on details of each scale respectively to obtain enhanced information, and then fuse the enhanced information of different scales to generate an analysis image.

[0122] In the embodiment of the present invention, the preprocessing module comprehensively optimizes the image quality through multiple processing of the noise suppression unit, the regional contrast enhancement unit, the edge refinement unit and the multi-scale detail processing unit. The noise suppression unit is responsible for removing noise interference in the image to ensure the clarity of the image, while retaining the key texture features of the ribbon surface, and providing high-quality image data for subsequent analysis. The regional contrast enhancement unit further analyzes the brightness difference in the noise suppression image, and improves the visibility of the subtle texture of the ribbon by adaptively adjusting the contrast, thereby more clearly showing the details of the ribbon surface.

[0123] The edge refinement unit enhances the edge of the image after regional contrast enhancement, generating a refined image with clear edge features, which significantly enhances the edge features of the ribbon surface. The multi-scale detail processing unit decomposes the refined image into multiple scales, further amplifies the image detail information through multiple scales of contrast enhancement, and makes the image have clear texture and edge effects at different levels. This layered enhancement process not only improves the macroscopic clarity of the image, but also improves the recognition of microscopic details, thereby ensuring the accuracy of subsequent defect detection.

[0124] Specifically, the noise suppression unit is used to detect and remove high-frequency noise in the corrected image while retaining the surface texture features of the ribbon to ensure the clarity of the image. The specific implementation includes: using edge-preserving noise reduction algorithms, such as bilateral filtering or non-local mean filtering, to reduce noise interference by identifying noisy pixels in the image and replacing or smoothing the noise points with neighborhood information. The output of this unit is an image with reduced noise and preserved details, providing purer image data for subsequent processing.

[0125] The regional contrast enhancement unit analyzes the local brightness difference of the image based on the noise suppression image, and enhances the contrast of subtle textures through adaptive contrast adjustment methods. For example, using histogram equalization or local contrast enhancement technology, the brightness of darker and brighter areas in the image is adjusted separately to make the details of the ribbon surface clearer. Regional contrast enhancement helps to better identify subtle defects on the ribbon during inspection.

[0126] The edge refinement unit is used to perform edge refinement processing on the image after regional contrast enhancement to make the edge of the ribbon texture sharper. Edge refinement uses edge enhancement algorithms, such as the Laplace operator or the Sobel operator, to make the edge appear clearer and sharper by calculating the second-order or high-order derivatives of the pixels at the edge, thereby retaining key edge information. This step can help subsequent defect detection to more accurately locate and identify boundaries.

[0127] The multi-scale detail processing unit decomposes the image into multiple scale levels and independently enhances the details of each scale level. Multi-scale image decomposition methods, such as wavelet decomposition or Laplace pyramid decomposition, are usually used to place different detail information of the image in different scale levels, and then enhance the contrast of each level. Finally, the detail information of each scale is fused to generate an enhanced analysis image. The processed image can present rich texture details at different scales, which is convenient for detecting defects of different sizes.

[0128] In a preferred embodiment of the present invention, the preliminary defect detection module includes:

[0129] The multi-level structure comparison unit is used to compare features between the analysis image and the preset comparison image, calculate the similarity, and identify the first area that deviates from the normal structure; the feature similarity calculation formula is as follows:

[0130] ,

[0131] in, For the coordinates The feature similarity at is the number of multi-scale layers, For the The weight of the layer scale, ,in is the weight coefficient, is the normalization term, and In the Analyzing images at the layer scale Compare image with preset The mean of and In the Analyzing images at the layer scale Compare image with preset The variance of To analyze the image Compare image with preset In the Covariance at the layer scale, and is a smoothing constant used to prevent the denominator from being zero;

[0132] The defect area marking unit is used to screen and mark the first area through a preset similarity threshold to generate a rough screening area image.

[0133] In an embodiment of the present invention, the preliminary defect detection module includes a multi-level structure comparison unit and a defect area marking unit to identify significant defect areas in the analysis image. The multi-level structure comparison unit first performs a multi-scale feature comparison between the analysis image and the preset comparison image, and calculates the similarity layer by layer to identify the area in the image that deviates from the normal structure. This multi-level feature comparison method not only takes into account the macroscopic image structure, but also can deeply analyze the image details, so as to effectively locate potential defects. After obtaining the preliminary screening area, the defect area marking unit marks the screening results by a preset similarity threshold, and only marks those areas that significantly deviate from the normal structure. This process enables the preliminary defect detection module to quickly and accurately locate the significant defect areas on the surface of the ribbon, thereby providing reliable basic data for subsequent fine analysis.

[0134] In the similarity calculation formula, Used to calculate the similarity between the analysis image and the preset comparison image at the k-th scale. and The analysis image and the comparison image are respectively The mean of the layer, For the two images in Covariance of the layers. It is used to smooth the calculation results and avoid the situation where the denominator is zero. Through this formula, the multi-level structure comparison unit can detect the image structure layer by layer at different scales to ensure the accuracy and robustness of preliminary defect detection.

[0135] Specifically, for the weight coefficient , can be automatically set according to the characteristics of the image. For example, the texture complexity, contrast or other image quality indicators of the image are used to determine the weights of different scale layers. For images with complex textures, the weight coefficients of higher scales will be higher to pay more attention to details; for simple patterns, the weight coefficients of lower scales will be higher to maintain the emphasis on the overall structure. It can also be determined through experimental optimization. In different ribbon images, the weight coefficients of different levels are adjusted through multiple experiments to observe their effects on defect detection effects. For example, at lower scale layers, more weights are given to focus on overall structural information; at higher scale layers, smaller weights are given to capture tiny details. Through experimental data analysis, the optimal weight distribution is selected.

[0136] In a preferred embodiment of the present invention, the fine defect analysis module includes:

[0137] A contour fitting unit is used to accurately fit the edge of the rough screening area image to generate a fine edge image;

[0138] The internal structure analysis unit is used to analyze the texture and shape characteristics of the edge fine image, identify the area containing tiny defects, and form a fine defect image.

[0139] In an embodiment of the present invention, based on the preliminary detection, the fine defect analysis module further refines the defect area through the contour fitting unit and the internal structure analysis unit, identifies and marks the tiny defects in the surface of the ribbon. The contour fitting unit accurately fits the edge of the defect area of ​​the coarse screening area image to generate a fine edge image. The fitting process can highly retain the true shape and boundary information of the defect, avoid the shape error caused by the preliminary detection of coarse screening, and ensure the accurate expression of the defect edge. The internal structure analysis unit identifies the potential internal tiny defect area based on the texture and shape features of the fine edge image. This deep analysis method can not only accurately distinguish the defect type, but also classify and mark its morphology, thereby improving the accuracy of defect recognition at the detail level.

[0140] Compared with the single-level or rough defect detection methods in the existing technology, this module ensures the integrity and precision of the surface defect information of the ribbon through the refinement process of edge fitting and structural analysis. It is particularly suitable for detecting those subtle defects that are difficult to detect, which helps to control and optimize the quality of ribbon production.

[0141] Specifically, the contour fitting unit accurately fits the defect edge of the rough screening area image to generate a fine edge image. The unit uses edge detection algorithms, such as the Canny algorithm or gradient detection method, to extract the edge contour of the defect, and further smoothes the defect edge through polynomial fitting, spline curve fitting and other methods. Through this fitting operation, the defect edge is made smoother and continuous, ensuring the accurate expression of the defect shape.

[0142] The internal structure analysis unit analyzes the internal texture and shape characteristics of the defect area based on the edge fine image after contour fitting. Use feature extraction algorithms such as local binary pattern LBP or gray level co-occurrence matrix GLCM to describe and analyze the internal texture characteristics of the defect area, thereby identifying potential defect areas with fine features. Through this analysis step, the system can conduct in-depth and detailed classification of the defect morphology, providing basic data for subsequent quality assessment.

[0143] In a preferred embodiment of the present invention, the adaptive filtering module includes:

[0144] A material property identification unit, used to automatically select a preset corresponding defect filtering parameter set according to the material type of the webbing;

[0145] The defect screening unit is used to screen each defect area in the fine defect image according to the defect filtering parameter set, and convert it into defect coordinate data of the defect area.

[0146] In an embodiment of the present invention, the adaptive filtering module realizes the defect information screening process for different webbing materials through the material characteristic recognition unit and the defect screening unit. The material characteristic recognition unit automatically identifies the material type of the webbing and automatically selects a defect filtering parameter set according to the characteristics of different materials. The recognition mechanism can adjust the screening criteria according to the specific material characteristics of the webbing, so that the system can maintain a high defect detection accuracy when facing different webbing materials. After acquiring the fine defect image, the defect screening unit screens each defect area according to the filtering parameter set selected by the material characteristic recognition unit, removes the misjudged noise area, and retains only the coordinate data of the valid defect area. This screening mechanism greatly reduces the noise interference caused by different materials and improves the accuracy of the defect data.

[0147] Through the material-adaptive defect screening function, this module is suitable for the detection needs of various webbing materials, ensuring the matching degree between the defect detection results and the webbing material, avoiding the misjudgment and missed detection problems caused by material changes in the existing technology, and significantly improving the versatility and accuracy of webbing quality detection.

[0148] Specifically, the material property recognition unit is responsible for automatically selecting a suitable defect filtering parameter set according to the material type of the webbing. The unit first identifies the material type of the webbing through camera data, image texture or a preset material parameter library, and selects suitable defect screening parameters for different materials based on the optical properties, surface features and other information of the material to ensure filtering accuracy. The parameters selected after material recognition can effectively reduce the interference of noise and false defects and improve detection accuracy.

[0149] The defect screening unit screens each defect area in the fine defect image based on the filter parameter set provided by the material characteristic recognition unit. The specific steps include: comparing the contrast, shape and texture characteristics of each defect area with the preset filter parameter set, eliminating pseudo-defect areas that do not meet the defect characteristics, and converting the remaining real defect areas into defect coordinate data. This can greatly reduce the false detection rate and make the final defect positioning more accurate.

[0150] In a preferred embodiment of the present invention, the dynamic coordinate mapping module includes:

[0151] The motion trajectory monitoring unit is used to monitor the motion trajectory of the webbing in the production process in real time and generate the position data of the webbing in the production process;

[0152] The dynamic position mapping unit is used to associate the defect coordinate data with the position data, determine the dynamic position of the defect in the physical space, and obtain the actual coordinate data.

[0153] In an embodiment of the present invention, the dynamic coordinate mapping module combines the defect coordinate data with the real-time motion trajectory of the webbing through a motion trajectory monitoring unit and a dynamic position mapping unit to achieve accurate positioning of the defect in physical space. The motion trajectory monitoring unit is used to monitor the motion trajectory of the webbing in real time during the production process, record the position information of the webbing at different time points, and generate corresponding motion data. The dynamic position mapping unit associates the defect coordinate data with the motion trajectory data, accurately maps the defect position to the physical coordinate system through dynamic position mapping, and generates actual coordinate data. This process ensures the synchronous association between the defect data and the webbing movement, and avoids the deviation of the defect position information caused by changes in the movement speed.

[0154] The application of this module effectively solves the problem of defect positioning deviation caused by webbing movement in the existing technology, ensuring real-time and accurate positioning of defects during the detection process. Regardless of how the webbing speed changes, the actual physical position of the defect can be accurately reflected in the final output, significantly improving the detection accuracy and the control level in the production process.

[0155] Specifically, the motion trajectory monitoring unit monitors the motion trajectory of the webbing in the production process in real time, records the actual position information of the webbing, and generates motion trajectory data. The speed sensors installed on both sides of the webbing measure and record the real-time position and speed changes of the webbing to ensure that the webbing position can be accurately tracked at different speeds. The role of this module is to provide dynamic position information of the webbing, providing an accurate position information basis for subsequent defect location.

[0156] The dynamic position mapping unit associates the defect coordinate data with the motion trajectory data to calculate the position of the defect in the physical space. The specific method is: during the movement of the webbing, the motion position information of the webbing is obtained in real time, and the detected defect coordinates are associated with the motion position to obtain the physical position of the defect in the actual coordinate system. This process keeps the defect data and the motion trajectory of the webbing synchronously associated, ensuring that the physical position of the defect can be accurately mapped in different motion states to achieve accurate positioning.

[0157] In a preferred embodiment of the present invention, the data storage and analysis module includes:

[0158] A historical data storage unit, used to record and store actual coordinate data and related test data to form a historical data set of webbing quality;

[0159] The data analysis unit performs statistical analysis on historical data sets, identifies the changing trends and quality issues of webbing quality, and generates quality reports and trend analysis reports.

[0160] In an embodiment of the present invention, the data storage and analysis module further improves the management and analysis capabilities of the webbing quality inspection system through the historical data storage unit and the data analysis unit. The historical data storage unit records and stores the actual coordinate data and related inspection data to form a historical data set of webbing quality. The data set contains webbing defect information over a long period of time, providing comprehensive data support for the optimization of the production process and quality trend analysis. The data analysis unit performs statistical analysis based on the historical data set to identify the changing trend of webbing quality and potential quality problems, thereby generating a quality report and a trend analysis report. This process enables the system to conduct in-depth analysis with the support of big data, and realizes the full automation and systematic nature of quality monitoring.

[0161] By combining historical defect data with quality analysis, the system can identify quality risks in advance and provide early warning information in trend reports, avoiding the lag of traditional manual monitoring mode and improving the foresight of quality control. This module is particularly suitable for production environments that require strict quality tracking and management, improving the overall quality level of webbing products and the intelligence of production.

[0162] Specifically, the historical data storage unit is used to record and store actual coordinate data and related test data to form a historical data set of webbing quality. The unit saves information such as defect location, defect type and test time of each test in the database to provide support for long-term webbing quality tracking and management. By storing and maintaining these historical data, a complete data source can be provided for subsequent quality trend analysis.

[0163] The data analysis unit analyzes historical data sets to identify the changing trends of webbing quality and potential quality problems. Statistical analysis and machine learning algorithms are used to process historical data. For example, the changing trends of webbing quality are analyzed by calculating the number of defects, distribution density and frequency, and quality reports and trend analysis reports are generated. Through these analysis results, the system can detect potential quality problems in the production process in advance, provide managers with real-time quality monitoring data, and improve the quality control capabilities of the production process.

[0164] In a preferred embodiment of the present invention, the alarm and control module includes:

[0165] A threshold determination unit is used to determine the number of defects and the rate of change in the trend analysis report, and to generate an alarm signal when the detected data reaches a preset alarm threshold;

[0166] A production adjustment unit, which is used to adjust the production parameters of the webbing according to the number of defects and the rate of change after receiving the alarm signal;

[0167] The automatic shutdown unit is used to automatically suspend the production process after the alarm signal is triggered for a preset number of times in succession.

[0168] In the embodiment of the present invention, the design of the alarm and control module realizes the automatic monitoring and dynamic regulation of abnormal quality conditions in the webbing production process through a multi-level response mechanism. First, the threshold determination unit makes a real-time determination of the number of defects and the rate of change in the trend analysis report. Once it is detected that these data reach the preset alarm threshold, the system will automatically generate an alarm signal, so that quality control can intervene at an early stage. Compared with the traditional quality management model, the present invention can automatically capture small changes in the number and trend of defects during the production process, effectively avoiding the risk of large-scale rework or production suspension due to the accumulation of quality problems, and improving the continuity and stability of production.

[0169] After receiving the alarm signal, the production adjustment unit automatically adjusts the production parameters of the webbing according to the number and change rate of defects in the alarm content, such as adjusting the running speed, tension or process parameters of the webbing, so as to reduce the probability of defects in time. In this way, when an alarm occurs, the system does not simply shut down for processing, but optimizes the production process through intelligent regulation of production parameters, realizing automated and flexible quality management. Compared with the existing manual adjustment method, the production adjustment unit can respond quickly when an abnormality occurs, effectively reducing the impact of quality problems on production efficiency.

[0170] The automatic shutdown unit further improves the safety and control of the system. When the system detects that the alarm signal is triggered for a preset number of times in a row, the automatic shutdown unit will suspend the production process to prevent the mass production of unqualified webbing. This mechanism ensures that when the system detects quality problems that cannot be improved by adjusting production parameters, production can be interrupted in a timely manner, avoiding the large-scale production of defective products from the root, and providing sufficient time for subsequent troubleshooting and problem solving. Compared with the traditional manual shutdown processing method, the automatic shutdown unit realizes the intelligent management of the production process, can respond quickly to quality abnormalities, ensure the stable operation of the production line, and improve the overall product quality pass rate.

[0171] An embodiment of the present invention further provides a method for detecting the quality of a webbing based on industrial camera detection, the method comprising the following steps:

[0172] Monitor the moving speed of the webbing in real time and generate synchronization signals;

[0173] Using multiple groups of industrial cameras to capture multi-angle images of the same position from different angles according to synchronization signals, initial image data is obtained;

[0174] Spatially align and fuse the initial image data to generate a high-resolution image;

[0175] Optically compensate the surface texture of the ribbon in the high-resolution image according to the angle difference of the collected light to obtain a corrected image;

[0176] De-noising and enhancing the corrected image to obtain an analysis image;

[0177] Perform preliminary defect detection on the analysis image according to the preset comparison image, preliminarily mark the significant areas that deviate from the normal structure, and obtain a rough screening area image;

[0178] Perform edge contour fitting on the coarse screening area to obtain the defect recognition area image, and analyze the internal texture and shape characteristics of the defect recognition area image to identify potential minor defects and generate a fine defect image;

[0179] According to the material type and surface characteristics of the ribbon, the fine defect image is filtered to obtain the defect coordinate data;

[0180] By dynamically associating the defect coordinate data with the motion trajectory of the ribbon, the physical position of the defect is accurately mapped to obtain the actual coordinate data;

[0181] Store and analyze historical actual coordinate data and related test data, and generate quality reports and trend analysis reports;

[0182] When the number of defects or the trend change rate in the trend analysis report reaches the preset threshold, an alarm signal is automatically issued, and production parameters are adjusted or production is suspended.

[0183] It should be noted that this method is a method corresponding to the above-mentioned system, and all implementation methods in the above-mentioned system embodiment are applicable to this embodiment and can achieve the same technical effect.

[0184] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. The ribbon quality inspection system based on industrial camera detection is characterized by: The system comprises: Speed ​​tracking module, used to monitor the moving speed of the webbing in real time and generate synchronization signals; The multi-angle acquisition module is used to capture multi-angle images of the same position from different angles through multiple groups of industrial cameras according to synchronization signals to obtain initial image data; An image fusion module is used to spatially align and fuse the initial image data to generate a high-resolution image; An optical correction module is used to optically compensate the surface texture of the ribbon in the high-resolution image according to the angle difference of the collected light to obtain a corrected image; A preprocessing module is used to perform denoising and enhancement processing on the corrected image to obtain an analysis image; A preliminary defect detection module is used to perform preliminary defect detection on the analysis image according to a preset comparison image, and to preliminarily mark significant areas that deviate from the normal structure to obtain a rough screening area image; The fine defect analysis module is used to fit the edge contour of the coarse screening area to obtain the defect recognition area image, analyze the internal texture and shape characteristics of the defect recognition area image, identify potential minor defects, and generate a fine defect image; Adaptive filtering module, used to filter fine defect images according to the material type and surface characteristics of the ribbon to obtain defect coordinate data; The dynamic coordinate mapping module is used to accurately map the physical position of the defect by dynamically associating the defect coordinate data with the motion trajectory of the ribbon to obtain the actual coordinate data; Data storage and analysis module, used to store and analyze historical actual coordinate data and related test data, and generate quality reports and trend analysis reports; The alarm and control module is used to automatically send out an alarm signal and adjust production parameters or suspend production when the number of defects or trend change rate in the trend analysis report reaches a preset threshold; The image fusion module comprises: A transformation matrix generation unit, used for obtaining a position vector and a rotation matrix of each industrial camera and calculating a space transformation matrix of each camera according to the position vector and the rotation matrix of each industrial camera; A spatial alignment unit, used for performing coordinate transformation on the initial image data according to the spatial transformation matrix to obtain an aligned image; The weighted fusion unit is used to fuse the aligned images according to the shooting angle and clarity of each industrial camera to obtain a high-resolution image. The calculation formula of the pixel value of the high-resolution image is: ， in, For high-resolution images at coordinates The pixel value at is the number of industrial cameras for image acquisition, For Camera The weight of image fusion, where ,in The camera shooting angle, is the camera’s resolution factor, is a normalization term used to ensure that the sum of the weights of all cameras is 1. For the Camera in The aligned image pixel value at position, is the fusion coefficient, , is the edge enhancement coefficient, which is used to balance the pixel value and edge information. For the The camera captures the image in Laplacian of position; in, , For the The standard deviation of the images taken by the cameras, is the weight coefficient, which is used to control the influence of camera clarity on edge clarity. For the The edge strength measure of the camera image, ,in and is the number of rows and columns of the image, and Respectively represent Camera images in and Direction gradient.

2. The ribbon quality detection system based on industrial camera detection according to claim 1 is characterized in that: The optical correction module comprises: An angle compensation unit is used to obtain the shooting angle data of each industrial camera, calculate the light deviation value at different angles according to the shooting angle data, and generate angle compensation parameters; The light correction unit is used to adjust the brightness value and color information of each pixel in the high-resolution image according to the angle compensation parameter to obtain a corrected image; wherein the calculation formula of the pixel value of the corrected image is: ， in, To correct the image in coordinates The pixel value at is the brightness compensation coefficient, is the incident angle of the light, is the color correction coefficient, is the angle influence coefficient, For shooting angle, is the edge correction coefficient, and The images are and The second partial derivative in the direction.

3. The ribbon quality detection system based on industrial camera detection according to claim 2 is characterized in that: The preprocessing module comprises: A noise suppression unit is used to detect and remove image noise in the modified image, reduce image noise interference through an edge-preserving noise reduction algorithm, and obtain a noise-suppressed image; A regional contrast enhancement unit is used to analyze the brightness difference and contrast of each pixel in the noise suppression image, improve the contrast of the subtle texture on the surface of the ribbon, and obtain a first enhanced image; An edge refinement unit, configured to refine the edge features of the ribbon texture using an edge enhancement algorithm according to the first enhanced image, so as to generate a second enhanced image; The multi-scale detail processing unit is used to decompose detail information of the second enhanced image at multiple scales, perform contrast enhancement processing on details of each scale respectively to obtain enhanced information, and then fuse the enhanced information of different scales to generate an analysis image.

4. The ribbon quality detection system based on industrial camera detection according to claim 3 is characterized in that: The preliminary defect detection module comprises: The multi-level structure comparison unit is used to compare features between the analysis image and the preset comparison image, calculate the similarity, and identify the first area that deviates from the normal structure; the feature similarity calculation formula is as follows: ， in, For the coordinates The feature similarity at is the number of multi-scale layers, For the The weight of the layer scale, ,in is the weight coefficient, is the normalization term, and In the Analyzing images at the layer scale Compare image with preset The mean of and In the Analyzing images at the layer scale Compare image with preset The variance of To analyze the image Compare image with preset In the Covariance at the layer scale, and is a smoothing constant used to prevent the denominator from being zero; The defect area marking unit is used to screen and mark the first area through a preset similarity threshold to generate a rough screening area image.

5. The ribbon quality detection system based on industrial camera detection according to claim 4 is characterized in that: The fine defect analysis module includes: A contour fitting unit is used to accurately fit the edge of the rough screening area image to generate a fine edge image; The internal structure analysis unit is used to analyze the texture and shape characteristics of the edge fine image, identify the area containing tiny defects, and form a fine defect image.

6. The ribbon quality detection system based on industrial camera detection according to claim 5 is characterized in that: The adaptive filtering module comprises: A material property identification unit, used to automatically select a preset corresponding defect filtering parameter set according to the material type of the webbing; The defect screening unit is used to screen each defect area in the fine defect image according to the defect filtering parameter set, and convert it into defect coordinate data of the defect area.

7. The ribbon quality detection system based on industrial camera detection according to claim 6 is characterized in that: The dynamic coordinate mapping module comprises: The motion trajectory monitoring unit is used to monitor the motion trajectory of the webbing in the production process in real time and generate the position data of the webbing in the production process; The dynamic position mapping unit is used to associate the defect coordinate data with the position data, determine the dynamic position of the defect in the physical space, and obtain the actual coordinate data.

8. The ribbon quality detection system based on industrial camera detection according to claim 7 is characterized in that: The data storage and analysis module includes: A historical data storage unit, used to record and store actual coordinate data and related test data to form a historical data set of webbing quality; The data analysis unit performs statistical analysis on historical data sets, identifies the changing trends and quality issues of webbing quality, and generates quality reports and trend analysis reports.

9. The ribbon quality detection method based on industrial camera detection is characterized in that: Applied to the system according to any one of claims 1 to 8, the method comprises the following steps: Monitor the moving speed of the webbing in real time and generate synchronization signals; Using multiple groups of industrial cameras to capture multi-angle images of the same position from different angles according to synchronization signals, initial image data is obtained; Spatially align and fuse the initial image data to generate a high-resolution image; Optically compensate the surface texture of the ribbon in the high-resolution image according to the angle difference of the collected light to obtain a corrected image; De-noising and enhancing the corrected image to obtain an analysis image; Perform preliminary defect detection on the analysis image according to the preset comparison image, preliminarily mark the significant areas that deviate from the normal structure, and obtain a rough screening area image; Perform edge contour fitting on the coarse screening area to obtain the defect recognition area image, analyze the internal texture and shape characteristics of the defect recognition area image, identify potential minor defects, and generate a fine defect image; According to the material type and surface characteristics of the ribbon, the fine defect image is filtered to obtain the defect coordinate data; By dynamically associating the defect coordinate data with the motion trajectory of the ribbon, the physical position of the defect is accurately mapped to obtain the actual coordinate data; Store and analyze historical actual coordinate data and related test data, and generate quality reports and trend analysis reports; When the number of defects or the trend change rate in the trend analysis report reaches the preset threshold, an alarm signal is automatically issued, and production parameters are adjusted or production is suspended.

Images