Visual detection system for inner side of belt blank weaving

Abstract

The invention provides an inner side visual detection system for weaving of a strip blank. The inner side visual detection system comprises an image acquisition module, an image processing module, a control module and an output module, the image acquisition module comprises at least one high-resolution camera, the high-resolution camera and a plurality of groups of adjustable light sources work cooperatively, and by adjusting the intensity, the angle and the direction of the light sources, the visual detection accuracy of the inner side visual detection system is improved. And the illumination uniformity in the image acquisition process is ensured, so that the influence of illumination variation on the image quality is reduced. According to the inner side visual inspection system for strip blank weaving, through an innovative image acquisition and processing technology, the problems that the image quality is unstable and the defect recognition capability is insufficient are remarkably solved. The image acquisition module adopts a high-resolution camera and an adjustable light source to cooperatively work, and the uniformity of images under different environment illumination is ensured by dynamically adjusting the intensity, angle and direction of the light source, so that the influence of external illumination change on the image quality is reduced.

Description

Technical Field

[0001] The invention relates to the technical field of visual inspection systems, in particular to an inner side visual inspection system for weaving a strip blank. Background Art

[0002] The visual inspection system for the inner side of the woven fabric is a detection device based on computer vision technology, which is mainly used to monitor the inner defects of the fabric during the woven fabric process. The basic structure of the system includes: image acquisition module, processing module, control module and output module. The image acquisition module is usually composed of a high-resolution camera and a light source, which is used to shoot the inner side of the woven fabric in real time. The position of the light source and the camera is adjusted according to the surface texture and lighting conditions of the fabric to ensure a clear image. The image processing module analyzes the collected images through computer vision algorithms to identify defects in the fabric, such as uneven weaving, yarn breakage, color difference, etc. The control module is responsible for coordinating the work of various hardware devices, and the output module displays the detection results, which can be subsequently processed through the screen or the automation system.

[0003] Although the system can effectively perform visual inspection of the inner side of the woven strip, it still has some defects. The image acquisition module may be affected by changes in external lighting, resulting in unstable image quality, which affects the accuracy of detection. The image processing algorithm has limited ability to identify complex fabric defects, especially in cases where the fabric texture is complex or the defects are relatively small, and misjudgment or omission may occur. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention provides a visual inspection system for the inner side of woven fabrics, which solves the problem that the image acquisition module may be affected by changes in external lighting, resulting in unstable image quality; and the problem that the image processing algorithm has limited ability to recognize complex fabric defects.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a visual inspection system for the inner side of a woven strip, comprising:

[0006] An image acquisition module, an image processing module, a control module, and an output module. The image acquisition module includes at least one high-resolution camera. The high-resolution camera works in conjunction with multiple groups of adjustable light sources to ensure uniform illumination during image acquisition by adjusting the intensity, angle, and direction of the light source, thereby reducing the impact of illumination changes on image quality.

[0007] The image processing module includes an image preprocessing unit, a defect recognition unit and a data analysis unit, which can automatically adjust image processing parameters under changing lighting conditions, optimize image quality and accurately identify defects on the inside of the fabric;

[0008] The control module coordinates various hardware modules through control signals to automatically start the image acquisition, processing and output process;

[0009] The output module includes a display unit and an alarm system, which displays the detection results and gives prompts through the alarm system when abnormalities are found.

[0010] Preferably, the image acquisition module adopts a dual-camera design, which is located at different positions of the fabric, so as to synchronously capture the inner and outer side images of the fabric and reduce the detection error caused by the camera angle problem.

[0011] Preferably, the image preprocessing unit includes a denoising algorithm, a contrast enhancement algorithm and an automatic white balance adjustment algorithm, which denoises the collected image, improves the contrast and automatically adjusts the illumination compensation to effectively reduce the impact of changes in external environmental illumination;

[0012] The denoising algorithm uses a mathematical formula based on mean filtering:

[0013]

[0014] Among them, I(x,y) represents the original image pixel value, I output (x, y) is the pixel value after denoising, k is the radius of the filter, and the defect recognition unit adopts an image classification network model based on deep learning. The model can accurately identify various defects on the inside of the fabric, including but not limited to uneven weaving, yarn breakage, and color difference, by training image data sets containing different types of defects. The deep learning model uses a convolutional neural network structure, and the specific structure is:

[0015] y=f(W n *f(W n-1 *…f(W1*x)))

[0016] Among them, y is the final output classification result, f(·) is the activation function, W1,W2,…,W n is the weight matrix of the convolutional layer, and x is the input image data.

[0017] Preferably, during the training process of the deep learning model, a reinforcement learning strategy is used to perform data amplification on small sample defects in the image, thereby improving the accuracy of the algorithm in complex fabric defect recognition; wherein the data amplification is performed by rotating, scaling, and cropping the image, using the formula:

[0018] I aug (x,y)=f(I(x,y),θ)

[0019] Among them, I aug(x, y) represents the amplified image, f(·) is the image transformation function, θ is the transformation parameter, and the image processing module also includes an adaptive illumination compensation algorithm, which dynamically adjusts the image brightness and contrast parameters according to the changes in the collected ambient light intensity during the detection process to improve the stability and recognition accuracy of the image;

[0020] The illumination compensation is achieved by using a weighted average algorithm:

[0021] I adjusted (x,y)=I(x,y)×α+I ambient ×(1-α)

[0022] Among them, I adjusted (x, y) is the adjusted image pixel, α is the adjustment factor, I ambient is the ambient light intensity.

[0023] Preferably, the image processing module adopts a multi-scale image analysis method, which can identify detail defects of fabrics at different resolutions and scales, and improve the detection capability of complex fabric defects. The multi-scale analysis is based on fractal theory, and its formula is:

[0024]

[0025] Among them, D is the fractal dimension, N(r) is the number of points when the scale is r, and r is the resolution scale.

[0026] Preferably, the defect recognition unit further includes an algorithm based on edge detection, which can extract edge information of the fabric surface and effectively detect defects such as yarn breakage and dislocation. The edge detection algorithm uses the Canny operator, and its formula is:

[0027]

[0028] Among them, G x With G y are the gradients of the image in the x and y directions respectively, and G(x,y) is the edge strength of the image.

[0029] Preferably, the control module exchanges data with the interface of the production line system, and obtains information on the fabric production speed and position during the detection process to ensure the real-time and accuracy of the detection. The data interaction adopts a time-series synchronization model:

[0030] T sync =T capture +T process

[0031] Among them, T sync is the synchronization time, T capture is the image acquisition time, Tprocess is the image processing time.

[0032] Preferably, the control module includes a self-diagnosis function module, which can automatically detect problems and give alarm information when the equipment fails or is abnormal, so that the operator can perform maintenance in time. The data analysis unit compares the image analysis results with historical data, generates a quality assessment report of the damaged woven fabric, and provides corresponding production optimization suggestions. During the defect identification process, the image processing module uses a regional growing algorithm to accurately locate the defective area, and combines cluster analysis to classify different defects, so as to improve the accuracy and efficiency of detection.

[0033] The present invention provides a visual inspection system for the inner side of a woven strip. It has the following beneficial effects:

[0034] The inner side visual inspection system of the woven fabric has significantly solved the problems of unstable image quality and insufficient defect recognition ability through innovative image acquisition and processing technology. The image acquisition module uses a high-resolution camera to work with an adjustable light source. By dynamically adjusting the intensity, angle and direction of the light source, it ensures the uniformity of the image under different ambient lighting conditions, thereby reducing the impact of external lighting changes on image quality. The dual-camera design enables the system to capture the inner and outer side images of the fabric at the same time, effectively eliminating the detection errors caused by viewing angle problems. In addition, the adaptive illumination compensation algorithm in the image processing module can adjust the image brightness and contrast in real time during the detection process, thereby ensuring clear images under complex lighting conditions.

[0035] The convolutional neural network model based on deep learning is used for fabric defect recognition, which can effectively identify various defects such as uneven weaving, yarn breakage, and color difference. Through enhanced learning and data amplification technology, the model can maintain high recognition accuracy when facing small sample defects, especially in the recognition of complex fabric defects. In addition, the image processing module introduces a multi-scale analysis method, which can identify detailed defects of fabrics at different resolutions and scales, further improving the detection capability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a schematic diagram of the structure of the present invention. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] like Figure 1 As shown, an embodiment of the present invention provides an inner side visual inspection system for a woven fabric, including an image acquisition module, an image processing module, a control module, and an output module. The image acquisition module adopts a dual-camera design, which is respectively located at different positions of the fabric to synchronously capture the inner and outer side images of the fabric, thereby reducing the detection error caused by the camera angle problem. The image processing module adopts a multi-scale image analysis method, which can identify the detail defects of the fabric from different resolutions and scales, and improve the detection capability of complex fabric defects. The multi-scale analysis is based on fractal theory, and its formula is:

[0039]

[0040] Among them, D is the fractal dimension, N(r) is the number of points when the scale is r, and r is the resolution scale.

[0041] The image acquisition module includes at least one high-resolution camera, which works together with multiple sets of adjustable light sources to ensure the uniformity of lighting during image acquisition by adjusting the intensity, angle and direction of the light source to reduce the impact of lighting changes on image quality.

[0042] The image processing module includes an image preprocessing unit, a defect recognition unit and a data analysis unit. It can automatically adjust image processing parameters under changing lighting conditions, optimize image quality and accurately identify defects on the inside of the fabric. The image preprocessing unit includes a denoising algorithm, a contrast enhancement algorithm and an automatic white balance adjustment algorithm. It denoises the collected images, improves the contrast and automatically adjusts the lighting compensation to effectively reduce the impact of changes in external environmental lighting.

[0043] The denoising algorithm uses a mathematical formula based on mean filtering:

[0044]

[0045] Among them, I(x,y) represents the original image pixel value, I output (x, y) is the pixel value after denoising, k is the radius of the filter, and the defect recognition unit adopts an image classification network model based on deep learning. The model can accurately identify various defects on the inside of the fabric, including but not limited to uneven weaving, yarn breakage, and color difference, by training image data sets containing different types of defects. The deep learning model uses a convolutional neural network structure, and the specific structure is:

[0046] y=f(W n *f(W n-1 *…f(W1*x)))

[0047] Among them, y is the final output classification result, f(·) is the activation function, W1,W2,…,W nis the weight matrix of the convolution layer, x is the input image data, and the defect recognition unit further includes an algorithm based on edge detection, which can extract edge information of the fabric surface and effectively detect defects such as yarn breakage and dislocation. The edge detection algorithm uses the Canny operator, and its formula is:

[0048]

[0049] Among them, G x With G y are the gradients of the image in the x and y directions respectively, and G(x,y) is the edge strength of the image.

[0050] The control module coordinates each hardware module through control signals and automatically starts the image acquisition, processing and output process. The control module exchanges data with the production line system through the interface. During the detection process, it communicates with the fabric production speed and position information to ensure the real-time and accuracy of the detection. The data interaction adopts the timing synchronization model:

[0051] T sync =T capture +T process

[0052] Among them, T sync is the synchronization time, T capture is the image acquisition time, T process is the image processing time.

[0053] The output module includes a display unit and an alarm system, which display the detection results and give prompts through the alarm system when abnormalities are found. During the training process of the deep learning model, an enhanced learning strategy is used to amplify the data of small sample defects in the image, thereby improving the accuracy of the algorithm in identifying complex fabric defects.

[0054] Among them, data augmentation is performed by rotating, scaling, and cropping the image using the formula:

[0055] I aug (x,y)=f(I(x,y),θ)

[0056] Among them, I aug (x, y) represents the amplified image, f(·) is the image transformation function, θ is the transformation parameter, and the image processing module also includes an adaptive illumination compensation algorithm, which dynamically adjusts the image brightness and contrast parameters according to the changes in the collected ambient light intensity during the detection process to improve the image stability and recognition accuracy.

[0057] Light compensation is achieved by using a weighted average algorithm:

[0058] I adjusted (x,y)=I(x,y)×α+Iambient ×(1-α)

[0059] Among them, I adjusted (x, y) is the adjusted image pixel, α is the adjustment factor, I ambient The control module includes a self-diagnosis function module, which can automatically detect problems and give alarm information when the equipment fails or is abnormal, so that the operator can perform maintenance in time. The data analysis unit compares the image analysis results with historical data to generate a quality assessment report on the damaged woven fabric and provide corresponding production optimization suggestions. During the defect identification process, the image processing module uses the regional growing algorithm to accurately locate the defective area, and combines cluster analysis to classify different defects to improve the detection accuracy and efficiency.

[0060] Table 1: Data table

[0061]

[0062] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A visual inspection system for the inner side of a woven strip, characterized in that: include: Image acquisition module, image processing module, control module, output module; The image acquisition module includes at least one high-resolution camera, which works in conjunction with multiple groups of adjustable light sources to ensure uniform illumination during image acquisition by adjusting the intensity, angle and direction of the light source, thereby reducing the impact of illumination changes on image quality; The image processing module includes an image preprocessing unit, a defect recognition unit and a data analysis unit, which can automatically adjust image processing parameters under changing lighting conditions, optimize image quality and accurately identify defects on the inside of the fabric; The control module coordinates various hardware modules through control signals to automatically start the image acquisition, processing and output process; The output module includes a display unit and an alarm system, which displays the detection results and gives prompts through the alarm system when abnormalities are found.

2. The inner side visual inspection system of the strip blank weaving according to claim 1 is characterized in that: The image acquisition module adopts a dual-camera design, which is located at different positions of the fabric to synchronously capture the inner and outer side images of the fabric, thereby reducing the detection error caused by the camera angle problem.

3. The inner side visual inspection system of a woven strip according to claim 1, characterized in that: The image preprocessing unit includes a denoising algorithm, a contrast enhancement algorithm and an automatic white balance adjustment algorithm, which denoises the collected image, improves the contrast and automatically adjusts the illumination compensation to effectively reduce the impact of changes in external environmental illumination; The denoising algorithm uses a mathematical formula based on mean filtering: Among them, I(x,y) represents the original image pixel value, I output (x, y) is the pixel value after denoising, k is the radius of the filter, and the defect recognition unit adopts an image classification network model based on deep learning. The model can accurately identify various defects on the inside of the fabric, including but not limited to uneven weaving, yarn breakage, and color difference, by training image data sets containing different types of defects. The deep learning model uses a convolutional neural network structure, and the specific structure is: y=f(W n *f(W n-1 *…f(W1*x))) Among them, y is the final output classification result, f(·) is the activation function, W1,W2,…,W n is the weight matrix of the convolutional layer, and x is the input image data.

4. The inner side visual inspection system of a woven strip according to claim 1, characterized in that: During the training process of the deep learning model, a reinforcement learning strategy is used to perform data amplification on small sample defects in the image, thereby improving the accuracy of the algorithm in complex fabric defect recognition; Among them, data augmentation is performed by rotating, scaling, and cropping the image using the formula: I aug (x,y)=f(I(x,y),θ) Among them, I aug (x, y) represents the amplified image, f(·) is the image transformation function, θ is the transformation parameter, and the image processing module also includes an adaptive illumination compensation algorithm, which dynamically adjusts the image brightness and contrast parameters according to the changes in the collected ambient light intensity during the detection process to improve the stability and recognition accuracy of the image; The illumination compensation is achieved by using a weighted average algorithm: I adjusted (x,y)=I(x,y)×α+I ambient ×(1-α) Among them, I adjusted (x, y) is the adjusted image pixel, α is the adjustment factor, I ambient is the ambient light intensity.

5. The inner side visual inspection system of a woven strip according to claim 1, characterized in that: The image processing module adopts a multi-scale image analysis method, which can identify the detail defects of the fabric from different resolutions and scales, and improve the detection ability of complex fabric defects. The multi-scale analysis is based on fractal theory, and its formula is: Among them, D is the fractal dimension, N(r) is the number of points when the scale is r, and r is the resolution scale.

6. The inner side visual inspection system of a woven strip according to claim 1, characterized in that: The defect recognition unit further includes an algorithm based on edge detection, which can extract edge information of the fabric surface and effectively detect defects such as yarn breakage and dislocation. The edge detection algorithm uses the Canny operator, and its formula is: Among them, G x With G y are the gradients of the image in the x and y directions respectively, and G(x,y) is the edge strength of the image.

7. The inner side visual inspection system of a woven strip according to claim 1, characterized in that: The control module exchanges data with the production line system interface, and obtains fabric production speed and position information during the detection process to ensure the real-time and accuracy of the detection. The data interaction adopts a time-series synchronization model: T sync =T capture +T process Among them, T sync is the synchronization time, T capture is the image acquisition time, T process is the image processing time.

8. The inner side visual inspection system of a woven strip according to claim 1, characterized in that: The control module includes a self-diagnosis function module, which can automatically detect problems and give alarm information when the equipment fails or is abnormal, so that the operator can perform maintenance in time. The data analysis unit compares the image analysis results with historical data, generates a quality assessment report of the damaged woven fabric, and provides corresponding production optimization suggestions. During the defect identification process, the image processing module uses a regional growing algorithm to accurately locate the defective area, and combines cluster analysis to classify different defects, so as to improve the accuracy and efficiency of detection.

Images