CN119147539B

CN119147539B - Ribbon quality inspection system and method based on industrial camera inspection

Info

Publication number: CN119147539B
Application number: CN202411624816.1A
Authority: CN
Inventors: 严伟淦; 蓝舟; 陶雨晴
Original assignee: Xiamen Qiute New Material Research Institute Co ltd
Current assignee: Xiamen Qiute New Material Research Institute Co ltd
Priority date: 2024-11-14
Filing date: 2024-11-14
Publication date: 2025-02-25
Anticipated expiration: 2044-11-14
Also published as: CN119147539A

Abstract

The invention provides a braid quality detection system and a method based on industrial camera detection, relating to the technical field of data processing, wherein the system comprises a speed tracking module; the system comprises 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, a self-adaptive filtering module, a dynamic coordinate mapping module, a data storage and analysis module, an alarm and control module and a control module, wherein the dynamic coordinate mapping module is used for mapping the physical positions of defects to obtain actual coordinate data, the data storage and analysis module is used for storing and analyzing historical actual coordinate data and related detection data to generate a quality report and a trend analysis report, the alarm and control module is used for automatically sending an alarm signal and adjusting production parameters or suspending production when the number of defects or the trend change rate in the trend analysis report reaches a preset threshold value.

Description

Ribbon quality detection system and method based on industrial camera detection

Technical Field

The invention relates to the technical field of data processing, in particular to a braid quality detection system and method based on industrial camera detection.

Background

The existing braid quality detection technology mainly relies on a manual visual detection mode, and an operator visually observes flaws and defects on the surface of the braid, such as fracture, burrs, chromatic aberration and the like. Although this method can find defects in the webbing to some extent, it relies on subjective judgment by an operator, and is susceptible to factors such as fatigue, distraction, and the like, resulting in insufficient stability and consistency of the detection results.

To address the deficiencies of manual inspection, the prior art has begun to incorporate automated visual inspection systems based on industrial cameras. The system collects the braid images through an industrial camera and automatically identifies and analyzes possible defects on the surface of the braid by combining an image processing algorithm, so that the detection speed and consistency are greatly improved. However, existing industrial camera-based detection systems still have some technical limitations. For example, in fast moving production lines, high speed movement of the webbing often results in blurring or distortion during image acquisition. Particularly, under the condition of complex textures or light reflection on the surface of the braid, the conventional image processing algorithm is difficult to effectively distinguish normal textures from defects, so that the accuracy of a detection result is affected.

Disclosure of Invention

The invention aims to provide a braid quality detection system and a method based on industrial camera detection, and aims to solve the problems in the background art.

In order to solve the technical problems, the technical scheme of the invention is as follows:

in a first aspect, a webbing quality detection system based on industrial camera detection, the system comprising:

The speed tracking module is used for monitoring the moving speed of the braid in real time and generating a synchronous signal;

The multi-angle acquisition module is used for capturing multi-angle images of the same position from different angles through a plurality of groups of industrial cameras according to the synchronous signals to obtain initial image data;

the image fusion module is used for carrying out space alignment on the initial image data and fusing the initial image data to generate a high-resolution image;

The optical correction module is used for carrying out optical compensation on the texture of the braid surface in the high-resolution image according to the angle difference of the collected light rays to obtain a corrected image;

the preprocessing module is used for denoising and enhancing the corrected image to obtain an analysis image;

The preliminary defect detection module is used for carrying out preliminary defect detection on the analysis image according to the preset contrast image, and carrying out preliminary marking on the salient region deviating from the normal structure to obtain a coarse screen region image;

The fine defect analysis module is used for carrying out edge contour fitting on the coarse screening area to obtain a defect identification area image, analyzing internal texture and shape characteristics of the defect identification area image, identifying potential micro defects and generating a fine defect image;

the self-adaptive filtering module is used for filtering the fine defect image according to the material type and the surface characteristics of the braid to obtain defect coordinate data;

The dynamic coordinate mapping module is used for accurately mapping the physical position of the defect by dynamically associating the defect coordinate data with the motion track of the braid so as to obtain actual coordinate data;

The data storage and analysis module is used for storing and analyzing the historical actual coordinate data and related detection data and generating a quality report and a trend analysis report;

and the alarm and control module is used for automatically sending an alarm signal when the number of defects in the trend analysis report or the trend change rate reaches a preset threshold value, and adjusting production parameters or suspending production.

Preferably, the image fusion module includes:

the transformation matrix generation unit is used for acquiring the position vector and the rotation matrix of each industrial camera and calculating the space transformation matrix of each camera according to the position vector and the rotation matrix of each industrial camera;

The space alignment unit is used for carrying out coordinate conversion on the initial image data according to the space transformation matrix to obtain an aligned image;

The weighting fusion unit is used for carrying out fusion processing on the aligned images according to the shooting angle and definition of each industrial camera to obtain high-resolution images, wherein the calculation formula of pixel values of the high-resolution images is as follows:

,

Wherein, In coordinates for high resolution imagesThe pixel value at which it is located,For the number of industrial cameras for image acquisition,Is a cameraWeights for image fusion, whereinWhereinFor the camera to take a picture of the angle,For the sharpness coefficient of the camera,To normalize the term, to ensure that the sum of the weights of all cameras is1,Is the firstThe camera is atThe aligned image pixel values at the locations,In order for the coefficient of fusion to be a function of,,For edge enhancement coefficients, for balancing pixel values with edge information,Is the firstThe camera shoots imagesA laplace operator of the location;

Wherein, ,Is the firstThe standard deviation of the captured image by the camera,Is a weight coefficient for controlling the influence weight of the camera definition on the edge definition,Is the firstEdge intensity measurements of the camera image of the table,WhereinAndIs the number of lines and columns of the image,AndRespectively represent the firstImage of cameraAndA gradient of direction.

Preferably, the optical correction module includes:

the angle compensation unit is used for acquiring shooting angle data of each industrial camera, calculating light deviation values under different angles according to the shooting angle data and generating angle compensation parameters;

The light correction unit is used for adjusting the brightness value and the 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 as follows:

,

Wherein, To correct the image at the coordinatesThe pixel value at which it is located,For the brightness compensation coefficient,For the angle of incidence of the light rays,As the color correction coefficient,As the angle-influencing factor,In order to take a picture of the angle,As the edge correction coefficient(s),AndRespectively the images are atAndSecond partial derivative of direction.

Preferably, the preprocessing module includes:

The noise suppression unit is used for detecting and removing image noise in the corrected image, reducing image noise interference through a noise reduction algorithm of edge preservation, and obtaining a noise suppression image;

The regional contrast improving unit is used for analyzing the brightness difference and contrast of each pixel in the noise suppression image and improving the contrast of fine textures on the surface of the braid to obtain a first improved image;

The edge refining unit is used for refining edge characteristics of the braid texture by adopting an edge enhancement algorithm according to the first lifting image to generate a second lifting image;

The multi-scale detail processing unit is used for decomposing detail information of the second lifting image on multiple scales, respectively carrying out contrast enhancement processing on the details of each scale to obtain enhancement information, and then fusing the enhancement information of different scales to generate an analysis image.

Preferably, the preliminary defect detection module includes:

The multi-layer structure comparison unit is used for comparing the characteristics of the analysis image with the preset comparison image, calculating the similarity to identify a first area deviating from a normal structure, wherein the similarity calculation formula of the characteristics is as follows:

,

Wherein, Is in coordinatesThe degree of similarity of the features at the location,For the number of layers in a multi-scale,Is the firstThe weight of the layer scale is calculated,WhereinAs the weight coefficient of the light-emitting diode,For the normalization of the terms,AndRespectively at the firstAnalysis of images at the layer scaleContrast image with presetIs used for the average value of (a),AndRespectively at the firstAnalysis of images at the layer scaleContrast image with presetIs a function of the variance of (a),To analyze imagesContrast image with presetIn the first placeCovariance at the layer scale is obtained,AndIs a smoothing constant for preventing denominator from being zero;

And the defect region marking unit is used for screening and marking the first region through a preset similarity threshold value to generate a coarse screening region image.

Preferably, the fine defect analysis module includes:

The contour fitting unit is used for precisely fitting the edges of the coarse screening area image to generate an edge fine image;

And the internal structure analysis unit is used for analyzing the texture and shape characteristics of the edge fine image, identifying the region containing the micro defects and forming a fine defect image.

Preferably, the adaptive filtering module comprises:

The material characteristic identification unit is used for automatically selecting a preset corresponding defect filtering parameter set according to the material type of the braid;

And the defect screening unit is used for screening each defect area in the fine defect image according to the defect filtering parameter set and converting the defect area into defect coordinate data of the defect area.

Preferably, the dynamic coordinate mapping module includes:

The motion track monitoring unit is used for monitoring the motion track of the braid in the production process in real time and generating position data of the braid in the production process;

And the dynamic position mapping unit is used for correlating the defect coordinate data with the position data, determining the dynamic position of the defect in the physical space and obtaining the actual coordinate data.

Preferably, the data storage and analysis module comprises:

the historical data storage unit is used for recording and storing actual coordinate data and related detection data to form a historical data set of the braid quality;

And the data analysis unit is used for carrying out statistical analysis on the historical data set, identifying the change trend and quality problem of the braid quality and generating a quality report and a trend analysis report.

In a second aspect, a method for detecting quality of a webbing based on industrial camera detection, the method comprising the steps of:

Monitoring the moving speed of the braid in real time, and generating a synchronous signal;

multiple groups of industrial cameras are used for capturing images at the same position from different angles according to the synchronous signals, so that initial image data are obtained;

performing space alignment on the initial image data and fusing the initial image data to generate a high-resolution image;

optical compensation is carried out on the texture of the surface of the braid in the high-resolution image according to the angle difference of the collected light rays, so as to obtain a corrected image;

denoising and enhancing the corrected image to obtain an analysis image;

Performing preliminary defect detection on the analysis image according to the preset contrast image, and performing preliminary marking on the salient region deviating from the normal structure to obtain a coarse screening region image;

performing edge contour fitting on the coarse screening area to obtain a defect identification area image, analyzing internal texture and shape characteristics of the defect identification area image, identifying potential micro defects and generating a fine defect image;

Filtering the fine defect image according to the material type and the surface characteristics of the braid to obtain defect coordinate data;

The physical position of the defect is accurately mapped by dynamically associating the defect coordinate data with the motion trail of the braid, so as to obtain actual coordinate data;

storing and analyzing historical actual coordinate data and related detection data to generate a quality report and a trend analysis report;

when the number of defects or the trend change rate in the trend analysis report reaches a preset threshold, an alarm signal is automatically sent out, and production parameters are adjusted or production is suspended.

The scheme of the invention at least comprises the following beneficial effects:

The system monitors the moving speed of the braid in real time through the speed tracking module and generates a synchronous signal for controlling a plurality of industrial cameras of the multi-angle acquisition module to synchronously work. The multi-angle acquisition module can acquire images on the surface of the braid from multiple angles under the guidance of the synchronous signals, so that the integrity of image capturing and the comprehensiveness of detailed information are ensured. The system solves the problem of information loss under the traditional single-angle acquisition, ensures that the system can still capture clear image data even under the condition of high-speed movement of the braid, and effectively reduces the possibility of image distortion and dislocation.

And carrying out spatial alignment and fusion on the initial image data acquired at multiple angles through an image fusion module to generate a high-resolution image. The space alignment ensures that images acquired at different angles are overlapped under the same coordinate system, and avoids image dislocation caused by shooting angle difference in the traditional system. The high-resolution image generated in the fusion process greatly enhances the image details, so that the subsequent detection module can perform defect analysis on a clearer data source, and the detection accuracy is improved. The optical correction module further optimizes the optical consistency of the images, eliminates the problems of brightness and color difference caused by different shooting angles through the angle compensation of light rays, obtains more uniform corrected images, and avoids misjudgment caused by the difference of light ray conditions.

In the preprocessing module, the system performs denoising and enhancement processing on the corrected image, and the generated analysis image has higher contrast and definition, so that a reliable image foundation is provided for the defect detection module. The preliminary defect detection module performs structural comparison by using the analysis image and a preset comparison image, performs preliminary marking on a salient region deviating from a normal structure, and generates a coarse screening region image. The process ensures the accuracy of preliminary screening and lays a foundation for subsequent fine analysis. The fine defect analysis module fits the edges of the defect area on the basis of coarse screening, and further analyzes texture and shape characteristics, so that potential micro defects are identified, and a fine defect image is generated. Through this degree of depth analysis, the system can realize accurate location and classification to various defects that the meshbelt surface exists, has greatly improved the accuracy and the carefully degree of detection.

The self-adaptive filtering module carries out targeted filtering on the fine defect image according to the material and surface characteristics of the braid, only the coordinate data conforming to the defect characteristics are reserved, and the interference caused by material noise is eliminated. The dynamic coordinate mapping module dynamically correlates the filtered defect coordinate data with the motion trail of the braid, accurately maps the position of the defect in the physical space, and ensures the traceability and positioning accuracy of the defect data in practical application. The data storage and analysis module stores and analyzes the actual coordinate data and the detection data to generate a historical trend report of the quality of the braid. The big data analysis mode can pre-judge quality problems in advance and provide precious decision basis for production management. When the system detects that the defect number or the trend change rate reaches a threshold value, the alarm and control module automatically triggers an alarm signal, and even adjusts production parameters or pauses production according to the requirement so as to timely control the diffusion of defects.

The multi-module arrangement of the system enables the system to have higher precision and automation level in detecting the quality of the mesh belt. Compared with the prior art, the system realizes more efficient defect identification and control in the whole process of acquisition, processing, analysis and feedback, and provides important guarantee for efficient production and quality management of the braid.

Drawings

Fig. 1 is a block diagram of a webbing quality detection system based on industrial camera detection according to an embodiment of the present invention.

Detailed Description

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

As shown in fig. 1, an embodiment of the present invention proposes a webbing quality detection system based on industrial camera detection, the system comprising:

In the embodiment of the invention, the system firstly monitors the moving speed of the braid in real time through the speed tracking module and generates a synchronous signal for controlling a plurality of industrial cameras of the multi-angle acquisition module to synchronously work. The multi-angle acquisition module can acquire images on the surface of the braid from multiple angles under the guidance of the synchronous signals, so that the integrity of image capturing and the comprehensiveness of detailed information are ensured. The system solves the problem of information loss under the traditional single-angle acquisition, ensures that the system can still capture clear image data even under the condition of high-speed movement of the braid, and effectively reduces the possibility of image distortion and dislocation.

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

,

In the embodiment of the invention, the image fusion module integrates the initial image data with high quality by using the transformation matrix generation unit, the space alignment unit and the weighted fusion unit. First, the transformation matrix generation unit generates a corresponding spatial transformation matrix by calculating a position vector and a rotation matrix of each industrial camera, thereby converting data acquired at multiple angles into the same coordinate space. The space alignment unit performs coordinate conversion on the initial image data of each camera by using a space transformation matrix, so that the data of all cameras can be accurately overlapped at the same position, and the processing mode effectively eliminates the deviation problem caused by multi-angle shooting. By the space alignment mode, the fused high-resolution image avoids distortion or dislocation, and lays a solid foundation for subsequent image processing.

In the weighted fusion unit, the system performs weighted fusion processing on the aligned images according to the shooting angle and definition of each industrial camera.Represent the firstThe weight adjustment result of the image of the camera after brightness adjustment and edge enhancement is achieved by adjusting parametersDifferent camera data may be equalized. The weighted fusion mode can more effectively highlight the details of the clear image, and meanwhile, the continuity of the image at the edge part is improved, so that the high-resolution image has higher quality under the integration of multi-angle information.

Specifically, the following is an acquisition method of each coefficient in the present embodiment:

For edge enhancement coefficients Can be adjusted by experimental determination of a series of test imagesValues to ensure that the fused image achieves the desired effect in terms of edge sharpness. For example by testing for differencesThe effect of the values on edge sharpening is optimized by comparing objective image quality metrics, such as peak signal to noise ratio PSNR or structural similarity SSIM. In addition, the edge intensity can be measured by an edge detection algorithm in image processing, such as Sobel operator or Canny edge detection, and the measurement result is used for dynamic adjustment. For example, the selection of the appropriate one is automatic in the fusion process based on the edge strengthValues.

For adjustment coefficientsA fixed one can be manually set according to empirical data of the performance and shooting distance of different camerasValues to ensure proper adjustment of sharpness coefficients. Can also gradually optimize through experiments on different definition requirementsValues to find a value that can achieve the best balance between edge sharpness and overall image quality. In addition, the experimental data can be automatically adjusted in a cross-validation mode, and the experimental data are different in terms ofEvaluating image quality under value, and finally selecting the best fusion effectValues.

Specifically, the transformation matrix generation unit is used for acquiring the position vector and the rotation matrix of each industrial camera, and calculating the spatial transformation matrix of each camera according to the position vector and the rotation matrix. Firstly, according to the installation position and direction of each camera, measuring and recording the three-dimensional position coordinates of the camera, namely position vectors, and the rotation information of the camera, namely a rotation matrix, and acquiring the data through a calibration tool or software. Then, combining the position vector and the rotation matrix, and constructing a space transformation matrix through homogeneous coordinate transformation, so that the image shot by each camera can be transformed into a unified three-dimensional space through the matrix. The space transformation matrix is used for subsequent image alignment, so that the data acquired at multiple angles can be unified under the same coordinate system.

The spatial alignment unit performs coordinate conversion on the initial image data using the spatial transformation matrix calculated by the transformation matrix generation unit to obtain an aligned image. The spatial alignment process converts 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 matrix transforming each pixel location. The alignment process can effectively eliminate deviation caused by the angle difference of multiple cameras, and lays a foundation for subsequent image fusion.

In a preferred embodiment of the present invention, the optical correction module includes:

,

In the embodiment of the invention, the image fusion module performs spatial alignment and fusion on the multi-angle images of the braid based on the initial image data to generate a high-resolution image. The space alignment is used for eliminating coordinate deviation caused by different shooting angles, so that data acquired by multiple angles can be unified under the same space coordinate system, and continuity among images is ensured. The high-resolution image not only enhances the details of the braid, but also enables the subsequent optical correction, preprocessing and defect detection module to process on the basis of a higher-quality data source, and compared with the single acquisition angle or asynchronous acquisition method in the prior art, the method effectively improves the detail presentation capability of detection. The generation of the high-resolution image provides more abundant information for subsequent image processing and defect analysis, and is beneficial to further improving the accuracy and stability of defect identification.

The optical correction module is used for effectively correcting the optical deviation of the high-resolution image under the cooperation of the angle compensation unit and the light correction unit. The angle compensation unit calculates an angle deviation value according to the shooting angle data of each camera, generates an angle compensation parameter, and further compensates the light ray difference of different angles. Therefore, the uniform optical performance of the surface texture of the ribbon in the high-resolution image can be ensured under different shooting angles, and the influence of brightness and color difference on the image is reduced. The light correction unit further adjusts brightness and color information of each pixel of the image based on the angle compensation parameters, ensures visual consistency of the high-resolution image, and provides a unified data base for subsequent analysis.

In the formula for the light ray correction,For combining brightness compensation coefficientsAnd color correction coefficientBased on incident angle of lightAnd (3) the non-linear compensation of brightness and color deviation is realized.Then by angle influence coefficientAnd edge correction coefficientComposition, brightness and color uniformity over different photographing angles and edge regions are ensured. The method can ensure the optical consistency of the high-resolution image under different light conditions, so that the optical compensation effect is more accurate.

for brightness compensation coefficients It can be determined by correcting for brightness variations in different ambient light. For example, in a dim light environment, a larger may be requiredValues to boost brightness. The brightness compensation effect can be adjusted by experiments to test different brightnessThe values are presented under various shooting conditions, so that an optimal value suitable for all scenes is found.

For color correction coefficientsCan be set according to the color deviation value actually collected during camera calibrationEnsuring that the colors in the image are as close to true as possible. The correction test can also be adjusted using a color correction card or a specific standard color model, such as sRGBAnd (3) ensuring that the color deviation under different light conditions is minimum.

Coefficient of influence on angleCan be adjusted step by step in the actual environmentTo observe its effect on image edge sharpness and brightness uniformity. Determination of suitability by Multi-Angle experimentsValues. The optical simulation software can also be used for simulating the incident angle and brightness attenuation of the light to obtain attenuation curves under different angles so as to determine the proper angleValues.

For edge correction coefficientsImage edge quality assessment indicators, such as gradient magnitude or contrast, may be used to setValues to balance edge sharpening effects and overall image smoothness. Also in a large amount of image data, the automatic adjustment can be realized by training a machine learning modelAnd (5) optimizing the edge enhancement effect.

Specifically, the main function of the angle compensation unit is to obtain shooting angle data of each industrial camera, calculate light ray deviation values under different angles based on the data, and generate corresponding angle compensation parameters so as to ensure optical consistency of the high-resolution image under different shooting angles. Firstly, the shooting angle of each camera is measured through an inclination angle sensor or a calibration system installed on the camera, and angle information is recorded. Then, according to the included angles between different shooting angles and the normal line of the surface of the mesh belt, the change of the incident angle of the light is calculated, and the brightness and color deviation caused by the angle are estimated. Finally, these deviation data are used to generate angle compensation parameters that are to be applied in subsequent light correction to effect adjustment of the brightness and color information in the image. Therefore, the optical deviation caused by the difference of shooting angles can be effectively reduced, and the consistency of the images in visual effect is ensured.

In a preferred embodiment of the present invention, the preprocessing module includes:

In the embodiment of the invention, the preprocessing module comprehensively optimizes the image quality through multiple processing of the noise suppression unit, the region contrast improving unit, the edge refining unit and the multi-scale detail processing unit. The noise suppression unit is responsible for removing noise interference in the image, ensuring the definition of the image, retaining key texture features on the surface of the braid, and providing high-quality image data for subsequent analysis. The regional contrast improving unit further analyzes the brightness difference in the noise suppression image, and improves the visibility of fine textures of the braid through self-adaptive adjustment of contrast, so that the details of the surface of the braid are displayed more clearly.

The edge thinning unit enhances the edges of the image after the region contrast is improved, and generates a thinned image with clear edge characteristics, so that the edge characteristics of the surface of the braid are obviously improved. The multi-scale detail processing unit carries out multi-scale decomposition processing on the refined image, and further amplifies detail information of the image through contrast enhancement of multiple scales, so that the image has clear textures and edge effects on different levels. The layering enhancement processing not only improves the macroscopic definition of the image, but also improves the identification degree of microscopic details, thereby ensuring the accuracy of subsequent defect detection.

Specifically, the noise suppression unit is used for detecting and removing high-frequency noise in the corrected image, and simultaneously retaining texture characteristics of the surface of the braid so as to ensure the definition of the image. Specific implementations include reducing noise interference by identifying noise pixels in the image and replacing or smoothing the noise points with neighborhood information using edge-preserving noise reduction algorithms, such as bilateral filtering or non-local mean filtering. The output of the unit is a noise reduced and detail preserved image, providing purer image data for subsequent processing.

The regional contrast improving unit analyzes local brightness difference of the image on the basis of the noise suppression image, and enhances the contrast of the fine texture through a self-adaptive contrast adjusting method. For example, using histogram equalization or local contrast enhancement techniques, the darker and lighter areas of the image are each adjusted in brightness to make the details of the webbing surface more clear. The area contrast enhancement helps to better identify fine defects on the ribbon upon inspection.

The edge thinning unit is used for performing edge thinning processing on the image with the region contrast improved, so that the edges of the texture of the ribbon are sharper. Edge refinement adopts an edge enhancement algorithm, such as a Laplacian operator or a Sobel operator, and the edge appears clearer and sharper by performing second-order or higher-order derivative calculation on pixels at the edge, so that key edge information is reserved. This step may help subsequent defect detection more accurately locate and identify boundaries.

The multi-scale detail processing unit performs independent enhancement processing on the detail of each scale layer by decomposing the image into a plurality of scale layers. A multiscale image decomposition method, such as wavelet decomposition or laplacian pyramid decomposition, is generally adopted, different detail information of an image is respectively placed on different scale layers, then contrast of each layer is enhanced, and finally the detail information of each scale is fused to generate an enhanced analysis image. The processed image can present rich texture details on different scales, so that defects of different sizes can be detected conveniently.

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

,

In the embodiment of the invention, the preliminary defect detection module comprises a multi-layer structure comparison unit and a defect area marking unit so as to identify a significant defect area in the analysis image. The multi-level structure comparison unit firstly carries out multi-scale feature comparison on the analysis image and the preset comparison image, and calculates the similarity layer by layer so as to identify the area deviating from the normal structure in the image. The multi-level characteristic comparison mode not only considers a macroscopic image structure, but also can deeply analyze image details, thereby effectively positioning potential defects. And after the preliminary screening area is obtained, the screening result is marked by the defect area marking unit through a preset similarity threshold value, and only areas which deviate from a normal structure remarkably are marked. This process enables the preliminary defect detection module to quickly and accurately locate significant defect areas of the web surface, providing reliable underlying data for subsequent fine analysis.

In the formula of the calculation of the similarity,Is used for calculating the similarity of the analysis image and the preset contrast image in the k layer scale,AndRespectively the analysis image and the contrast image are in the firstThe average value of the layers is calculated,For two images at the firstCovariance of layers.And the method is used for smoothing the calculation result and avoiding the situation that the denominator is zero. Through the formula, the multi-layer structure comparison unit can detect the image structure layer by layer on different scales, and the accuracy and the robustness of preliminary defect detection are ensured.

Specifically, for the weight coefficientThe setting may be automatic according to the characteristics of the image. For example, the texture complexity, contrast, or other image quality index of the image is used to determine the weights of the different scale layers. For images with complex textures, the higher scale weight coefficients will be higher to focus more on details, and for simple patterns, the lower scale weight coefficients will be higher to maintain focus on the overall structure. It can also be determined by experimental optimisation. In different braid images, the influence of the weight coefficients of different layers on the defect detection effect is observed through multiple experiments. For example, at lower scale levels more weight is given to focus on overall structural information, and at higher scale levels less weight is given to capture minutiae. And (5) selecting optimal weight distribution through experimental data analysis.

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

In the embodiment of the invention, the fine defect analysis module further refines the defect area through the contour fitting unit and the internal structure analysis unit on the basis of preliminary detection, and identifies and marks the micro defects in the surface of the braid. And the contour fitting unit is used for carrying out accurate fitting on the edges of the defect areas of the coarse screen area image to generate an edge fine image. The fitting process can highly reserve the real shape and boundary information of the defect, avoid shape errors caused by preliminary detection of coarse screening, and ensure accurate expression of the edge of the defect. The internal structure analysis unit identifies an internal potential micro defect area based on texture and shape features of the edge fine image. The depth analysis mode not only can accurately distinguish the defect types, but also can carry out classification marking on the forms of the defects, thereby improving the accuracy of defect identification in the detail level.

Compared with the single-layer or rough defect detection mode in the prior art, the module ensures the integrity and fineness of the defect information on the surface of the braid through the refinement process of edge fitting and structure analysis, is particularly suitable for detecting the tiny defects which are difficult to detect, and is beneficial to quality control and optimization of braid production.

Specifically, the contour fitting unit performs accurate fitting on the defective edges of the coarse screen area image to generate an edge fine image. The unit extracts the edge contour of the defect by using an edge detection algorithm, such as a Canny algorithm or a gradient detection method, and further performs smoothing treatment on the edge of the defect by using a polynomial fitting method, a spline curve fitting method and the like. Through the fitting operation, the defect edge is smoother and more continuous, and accurate expression of the defect shape is ensured.

The internal structure analysis unit analyzes internal texture and shape characteristics of the defective region based on the edge fine image after contour fitting. The internal texture features of the defect region are described and analyzed using a feature extraction algorithm, such as a local binary pattern LBP or a gray level co-occurrence matrix GLCM, to identify potential defect regions having fine features. Through the analysis step, the system can conduct deep refinement classification on the form of the defect, and basic data is provided for subsequent quality assessment.

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

In the embodiment of the invention, the self-adaptive filtering module realizes the defect information screening processing aiming at different braid materials through the material characteristic identification unit and the defect screening unit. The material characteristic identification unit automatically identifies the material type of the braid and automatically selects a defect filtering parameter set according to the characteristics of different materials. The identification mechanism can adjust the screening standard according to specific material characteristics of the braid, so that the system can keep higher defect detection accuracy when facing different braid materials. And after the fine defect image is obtained, the defect screening unit screens each defect area according to the filtering parameter set selected by the material characteristic identification unit, removes the misjudged noise area and only retains the coordinate data of the effective defect area. The screening mechanism greatly reduces noise interference caused by different materials and improves the accuracy of defect data.

Through the defect screening function of material self-adaptation, this module is applicable to the detection demand of multiple meshbelt material, ensures the matching degree of defect testing result and meshbelt material, has avoided misjudgement and the omission problem that leads to because of the material changes among the prior art, has showing commonality and the precision that has improved meshbelt quality testing.

Specifically, the texture characteristic recognition unit is responsible for automatically selecting a proper defect filtering parameter set according to the texture type of the webbing. The unit firstly identifies the type of the material of the braid through the shooting data, the image texture or a preset material parameter library, and selects proper defect screening parameters for different materials by combining the information such as the optical characteristics, the surface characteristics and the like of the material so as to ensure the filtering precision. The parameters selected after the material identification can effectively reduce noise and interference of pseudo defects, and improve detection accuracy.

The defect screening unit screens each defect area in the fine defect image based on the filtering parameter set provided by the material characteristic identifying unit. The specific steps include comparing the contrast, shape and texture characteristics of each defect area with a preset filtering parameter set, eliminating the pseudo defect areas which do not accord with the defect characteristics, and converting the residual real defect areas into defect coordinate data. Therefore, the false detection rate can be greatly reduced, and the final defect positioning is more accurate.

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

In the embodiment of the invention, the dynamic coordinate mapping module combines the defect coordinate data with the real-time motion track of the braid through the motion track monitoring unit and the dynamic position mapping unit, so as to realize the accurate positioning of the defect in the physical space. The motion track monitoring unit is used for monitoring the motion track of the braid in the production process in real time, recording the position information of the braid at different time points and generating corresponding motion data. The dynamic position mapping unit correlates the defect coordinate data with the motion trail data, precisely maps the defect position into a physical coordinate system through dynamic position mapping, and generates actual coordinate data. This process ensures a synchronous correlation between the defect data and the web motion, avoiding deviation of defect position information due to change of motion speed.

The application of the module effectively solves the problem of defect positioning deviation caused by webbing movement in the prior art, and ensures the real-time accurate positioning of defects in the detection process. Regardless of the change of the speed of the webbing, the actual physical position of the defect can be accurately reflected in the final output, and the detection precision and the control level in the production process are remarkably improved.

Specifically, the motion trail monitoring unit monitors the motion trail of the braid in the production process in real time, records the actual position information of the braid, and generates motion trail data. Real-time position and speed change of the webbing are measured and recorded through speed sensors arranged on two sides of the webbing, so that accurate tracking of the webbing position at different speeds is ensured. The function of this module is to provide dynamic position information of the webbing, providing an accurate basis for position information for subsequent defect localization.

The dynamic position mapping unit correlates the defect coordinate data with the motion trail data to calculate the position of the defect in the physical space. The method comprises the steps of acquiring moving position information of a braid in real time in the moving process of the braid, and associating detected defect coordinates with the moving position to obtain the physical position of the defect under an actual coordinate system. The process ensures that the defect data and the motion trail of the braid keep synchronous association, ensures that the physical position of the defect can be accurately mapped under different motion states, and realizes accurate positioning.

In a preferred embodiment of the present invention, the data storage and analysis module comprises:

In the embodiment of the invention, the data storage and analysis module further improves the management and analysis capacity of the braid quality detection system through the historical data storage unit and the data analysis unit. The history data storage unit records and stores actual coordinate data and related detection data to form a history data set of the webbing quality. The data set contains ribbon defect information with long time span, and provides comprehensive data support for optimization and quality trend analysis of the production process. The data analysis unit performs statistical analysis based on the historical data set, and identifies a change trend of the web quality and a potential quality problem, thereby generating a quality report and a trend analysis report. The process enables the system to conduct deep analysis under the support of big data, and achieves comprehensive automation and systematicness of quality monitoring.

By combining the historical defect data with the quality analysis, the system can identify the quality risk in advance and provide early warning information in the trend report, so that the hysteresis of the traditional manual monitoring mode is avoided, and the prospective of quality control is improved. The module is particularly suitable for production environments requiring strict quality tracking and management, and improves the overall quality level of the braid product and the intellectualization of production.

Specifically, the history data storage unit is used for recording and storing actual coordinate data and related detection data to form a history data set of the webbing quality. The unit stores information such as defect positions, defect types, detection time and the like detected each time in a database, and provides support for long-time ribbon quality tracking and management. By storing and maintaining this historical data, a complete source of data can be provided for subsequent quality trend analysis.

The data analysis unit analyzes the historical data set and identifies the change trend of the quality of the braid and potential quality problems. The historical data is processed by adopting statistical analysis and machine learning algorithms, for example, the change trend of the quality of the braid is analyzed by calculating the number of defects, distribution density and frequency, and a quality report and a trend analysis report are generated. Through the analysis results, the system can discover potential quality problems in the production process in advance, provide real-time quality monitoring data for a manager, and improve the quality control capability of the production process.

In a preferred embodiment of the present invention, the alarm and control module comprises:

the threshold value judging unit is used for judging the number of defects and the change rate in the trend analysis report, and generating an alarm signal when the data reach a preset alarm threshold value;

The production adjusting unit is used for adjusting production parameters of the braid according to the defect number and the change rate after receiving the alarm signal;

And the automatic stopping unit is used for automatically stopping the production process after the number of times of triggering the alarm signal continuously reaches the preset number of times.

In the embodiment of the invention, the design of the alarm and control module realizes the automatic monitoring and dynamic regulation of the quality abnormal condition in the webbing production process through a multistage response mechanism. Firstly, the threshold value judging unit judges the number of defects and the change rate in the trend analysis report in real time, and once the data are detected to reach a preset alarm threshold value, the system automatically generates an alarm signal so that the quality control can intervene in an early stage. Compared with the traditional quality management mode, the invention can automatically capture the small changes of the defect quantity and trend in the production process, effectively avoid the risk of large-scale reworking or production stopping caused by accumulation of quality problems, and promote the continuity and stability of production.

After receiving the alarm signal, the production adjustment unit automatically adjusts production parameters of the braid, such as operation speed, tension or technological process parameters of the braid, according to the number and change rate of defects in alarm content, so as to reduce the probability of defects in time. Therefore, when an alarm occurs, the system does not simply stop processing, but optimizes the production process through intelligent regulation and control of production parameters, and automatic and flexible quality management is realized. Compared with the existing manual adjustment mode, the production adjustment unit can quickly respond when abnormality occurs, and the influence of quality problems on production efficiency is effectively reduced.

The automatic stopping unit further improves the safety and control strength of the system. When the system detects that the number of times of triggering the alarm signal continuously reaches the preset number of times, the automatic stopping unit can pause the production process, and mass production of unqualified woven belts is prevented. The mechanism ensures that when the system detects the quality problem which cannot be improved by adjusting the production parameters, the production can be interrupted in time, so that the large-scale production of defective products is radically avoided, and sufficient time is provided for subsequent fault detection and problem solving. Compared with the traditional manual shutdown processing mode, the automatic shutdown unit realizes intelligent management of the production process, can quickly react under the condition of abnormal quality, ensures stable operation of the production line, and improves the quality qualification rate of the whole product.

The embodiment of the invention also provides a braid quality detection method based on industrial camera detection, which comprises the following steps:

denoising and enhancing the corrected image to obtain an analysis image;

It should be noted that, the method is a method corresponding to the above system, and all implementation manners in the above system embodiment are applicable to the embodiment, so that the same technical effects can be achieved.

While the foregoing is directed to the preferred embodiments of the present invention, it will be appreciated by those skilled in the art that various modifications and adaptations can be made without departing from the principles of the present invention, and such modifications and adaptations are intended to be comprehended within the scope of the present invention.

Claims

1. Ribbon quality detection system based on industrial camera detection, characterized in that it comprises:

the alarm and control module is used for automatically sending out an alarm signal when the number of defects or the trend change rate in the trend analysis report reaches a preset threshold value, and adjusting production parameters or suspending production;

the image fusion module comprises:

,

2. The web quality detection system based on industrial camera detection of claim 1, wherein the optical correction module comprises:

,

3. The web quality detection system based on industrial camera detection of claim 2, wherein the preprocessing module comprises:

4. A web quality inspection system based on industrial camera inspection as claimed in claim 3 wherein the preliminary defect detection module comprises:

,

5. The web quality inspection system based on industrial camera inspection of claim 4, wherein the fine defect analysis module comprises:

6. The industrial camera detection-based webbing quality detection system of claim 5, wherein the adaptive filtering module comprises:

7. The industrial camera detection-based webbing quality detection system of claim 6, wherein the dynamic coordinate mapping module comprises:

8. The web quality detection system based on industrial camera detection of claim 7, wherein the data storage and analysis module comprises:

9. A web quality detection method based on industrial camera detection, characterized by being applied in a system according to any one of claims 1 to 8, the method comprising the steps of:

denoising and enhancing the corrected image to obtain an analysis image;