A belt conveyor surface tearing detection method based on image processing

Abstract

The application relates to a belt conveyor surface tearing detection method based on image processing, which comprises the following steps: setting hardware, obtaining a mask through reference images, processing the to-be-detected images by using the mask, and analyzing whether defects exist in the to-be-detected images. The belt conveyor surface tearing detection method based on image processing can identify various defect types such as cracks and scratches in real time. Compared with a traditional manual detection method, the adaptive system can process a large amount of data in a shorter time, reduces human errors, and improves the reliability of detection.

Description

Technical Field

[0001] This invention relates to a method for detecting surface tearing on a belt conveyor based on image processing, belonging to the field of electrical equipment technology. Background Technology

[0002] Traditional belt tear detection methods mainly include mechanical switch detection and machine vision detection. Mechanical switch detection involves installing a tear protection switch under the belt; when the belt tears, material leaks out, triggering the switch and issuing an alarm. Machine vision detection methods use single or multiple cameras installed at different locations on the belt to analyze images and determine if a tear has occurred. Currently, the most common machine vision method for belt tear detection is bottom detection, which offers the advantage of real-time monitoring of the belt's condition and integration with control systems such as PLCs, achieving a high degree of automation and reducing manual intervention. However, this method also has some drawbacks. First, bottom detection can only monitor for through-tears or material leakage, failing to detect potential defects on the material conveying surface in a timely manner, potentially leading to early damage being overlooked. Second, this approach typically relies on alarm systems, which may only react after a tear has occurred, causing a response delay and impacting overall production efficiency. Therefore, although traditional detection methods ensure production safety to some extent, their limitations remain significant, necessitating the exploration of more advanced detection technologies.

[0003] To detect initial defects on the material transport surface, a camera installed directly above the belt can be used to capture images of the belt when no material is being transported. However, due to the complex texture of the belt surface and the possibility of material residue, this poses a significant challenge to subsequent image processing algorithms and greatly affects real-time performance. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the shortcomings of the above-mentioned technologies and provide a machine vision method for detecting belt defects based on laser projection deformation analysis.

[0005] To solve the above-mentioned technical problems, the technical solution proposed by this invention is: a method for detecting surface tearing of a belt conveyor based on image processing, comprising the following steps: Step 1: Set up a laser and a camera above the belt and then capture images. The laser is directed towards the belt and the laser beam can cover the width of the belt. The camera can capture images of the areas on the belt that are illuminated by the laser. Step 2: Input a reference image of a belt without defects, perform grayscale thresholding on the reference image, and extract the regions where the grayscale value exceeds the grayscale threshold; Step 3: Use morphological dilation to connect adjacent regions, generating a ROI region larger than the actual laser projection line; Step 4: Perform connected component analysis to identify the largest connected region and generate a Boolean mask; Step 5: Acquire the image to be detected captured by the camera, and extract the ROI region image using a Boolean mask; Step 6: Calculate the dynamic threshold of each column in the ROI region image and generate a binary image; Step 7: Divide the binary image into several columns and calculate the cumulative pixel sum for each group; Step 8: Calculate the difference of the cumulative sum of pixels in each group. The position where the difference is greater than 0 indicates that there is a defect on the belt corresponding to that group of pixels.

[0006] A further improvement to the above scheme is that: the specific method of step 6 is to calculate the pixel value of each column in the ROI region image, calculate the average value of pixels whose normalized pixel gray value is less than the gray value threshold, and add an offset as the dynamic threshold of the corresponding column; then compare the pixel value of the column with the calculated dynamic threshold to generate a binarization result and obtain a binary image.

[0007] A further improvement to the above scheme is that the grayscale threshold is 0.9 and the offset is 0.2.

[0008] A further improvement to the above scheme is that, in step 5, after acquiring the image to be detected captured by the camera, median filtering is first performed.

[0009] A further improvement to the above scheme is that, in step 7, median filtering is performed before grouping the binary images.

[0010] A further improvement to the above scheme is that, in step 2, the grayscale threshold is 90% of the highest value.

[0011] A further improvement to the above scheme is that the laser emitted by the laser is elongated.

[0012] The image processing-based surface tear detection method for conveyor belts provided by this invention can identify various defect types such as cracks and scratches in real time. Compared with traditional manual inspection methods, this adaptive system can process large amounts of data in a shorter time, reduce human error, and improve the reliability of detection.

[0013] This invention provides an image processing-based method for detecting surface tears in belt conveyors. By acquiring laser projection line images of the belt conveyor surface, compared to belt bottom detection methods, it can detect surface defects more promptly, allowing for early intervention and preventing severe belt tears. It achieves real-time belt defect detection: the proposed adaptive binarization and cumulative column summation calculation process ensures efficient belt defect detection, suitable for industrial production and other scenarios requiring real-time monitoring. Anti-interference capability is improved: by using laser projection line deformation analysis and fusing image processing technology, the impact of noise and belt surface texture interference on the detection results is effectively reduced, achieving the goal of detecting belt surface defects. Reduced need for manual intervention: through automated detection, reliance on manual inspection is reduced, lowering labor costs and operational risks. Attached Figure Description

[0014] Figure 1 This is a reference diagram used in a preferred embodiment of the present invention.

[0015] Figure 2 This is a reference diagram of the ROI region obtained in a preferred embodiment of the present invention.

[0016] Figure 3 This is the image to be detected in a preferred embodiment of the present invention.

[0017] Figure 4 This is a summation diagram in a preferred embodiment of the present invention.

[0018] Figure 5 This is a schematic diagram of the results of belt breakage defect detection.

[0019] Figure 6 This is a schematic diagram of the detection results of belt damage defects. Detailed Implementation

[0020] Example: The image processing-based belt conveyor surface tear detection method of this example includes the following steps: Step 1: Set up a laser and a camera above the belt, and then capture images. The laser emitted by the laser is directed towards the belt and is long and can cover the width of the belt. The camera can capture images of the areas on the belt that are illuminated by the laser.

[0021] Step 2: Input a reference image of a defect-free belt. Perform grayscale thresholding on the reference image and extract areas where the grayscale value exceeds the threshold. In this embodiment, the grayscale threshold is 90%, which is 255*980%, to extract areas where the grayscale value exceeds 90%. The reference image is shown below. Figure 1 As shown.

[0022] Step 3: Use morphological dilation to connect adjacent regions, generating a Region of Interest (ROI) larger than the actual laser projection line; this improves image detection efficiency and facilitates reliable extraction of the actual laser projection line. The ROI region is shown below. Figure 2 As shown.

[0023] Step 4: Perform connected component analysis to identify the largest connected region and generate a Boolean mask.

[0024] After the above steps, a Boolean mask for processing the image to be detected is obtained. This greatly reduces the influence of complex background images at other locations on the belt surface on the detection results, and also significantly reduces the number of pixels required for subsequent image processing.

[0025] Step 5: Acquire the image to be detected captured by the camera, apply median filtering to reduce noise interference in the image and improve the effect of subsequent processing; then use a Boolean mask to extract the ROI region image. The image to be detected is shown in Figure 3.

[0026] Step 6: Calculate the dynamic threshold of each column in the ROI region image and generate a binary image.

[0027] Specifically, for each column of the ROI region image, the pixel values ​​are calculated. The average value of pixels with normalized grayscale values ​​less than a grayscale threshold is calculated, and an offset is added to it as the dynamic threshold for the corresponding column. Then, the pixel values ​​of the column are compared with the calculated dynamic threshold to generate a binarized result, resulting in a binary image. In this embodiment, a grayscale threshold of 0.9 and an offset of 0.2 are used.

[0028] The surface of conveyor belts typically has vertical stripes. The reflection of laser lines varies at different locations on the belt surface, resulting in differences in the grayscale values ​​of the laser line images at different locations. Using a global thresholding method or binarization based on the grayscale features of the image within a square neighborhood will lead to problems with the binarization results. Considering the vertical stripe feature of the belt and the differences in image grayscale values ​​at different locations, this embodiment binarizes each column of pixels within the ROI region individually. The binarization threshold is the average value of the corresponding column within the ROI region plus a bias amount.

[0029] Step 7: Perform median filtering on the binary image to remove isolated noise points and further sharpen the image. Divide the binary image into several columns and calculate the pixel sum of each group. In this embodiment, 5 columns are used as a group, and the image is plotted as follows: Figure 4 As shown, the horizontal axis represents the group number, and the vertical axis represents the cumulative sum.

[0030] Step 8: Calculate the difference between the cumulative sums of pixels in each group. Positions with a difference greater than 0 indicate a defect in the belt corresponding to that group of pixels. Figure 4 The boxed area in the middle.

[0031] By calculating the column sums of the processed image, summing each column of the binary image, the active regions present in each column can be effectively evaluated, and the total number of active pixels (value 1) can be calculated, thus generating a column sum array. This process not only effectively evaluates the active regions present in each column but also quickly identifies potential defect locations. Specifically, firstly, the total number of active pixels in each column is calculated, generating a column sum array that visually reflects the distribution of active regions in each column. Through analysis of the column sum array, it is possible to quickly identify which columns have a large number of active pixels; these columns may correspond to defect areas in the image. Further introducing difference analysis, by calculating the difference between adjacent column sums, columns with significant changes can be identified. This process helps highlight areas that show obvious changes compared to surrounding columns, and these areas are usually more likely to be defective. Combining the column sum calculation results and difference analysis, automatic identification of defect areas can be achieved, thereby significantly reducing the need for manual intervention and improving the efficiency and accuracy of detection. By using the above steps, the calculation of column sums and row sums, as well as difference analysis, we can not only effectively evaluate the active regions in the image, but also provide a reliable basis for the automatic identification of defects, thereby achieving more efficient quality control in industrial applications.

[0032] Furthermore, this embodiment also includes: Create a graphics window; use Matplotlib to create a graphics window with subplots to display the processing results at each stage. If defects are detected, mark the defect areas with red rectangles on the "Detected image" subplot to provide a visual representation of the results. See [link to Matplotlib]. Figure 5 and Figure 6 The two different defects shown serve as warnings for belt defects, alerting workers to take precautions to reduce potential safety hazards and production losses.

[0033] This embodiment of the image processing-based belt conveyor surface defect detection method aims to improve detection accuracy and efficiency. The method acquires belt images in real time and determines the region of interest (ROI). It utilizes line laser illumination technology to overcome the influence of complex textures and material residues, employs column-by-column dynamic threshold binarization to extract the target region, and removes noise through median filtering, ultimately achieving effective identification of belt surface defects. This method not only monitors potential defects in real time, avoiding the response delay of traditional methods, but also enhances the diversity of training data by simulating defects, improving the model's generalization ability, thus showing broad prospects in industrial applications. In the future, with continuous technological improvement and the expansion of application scenarios, this system will provide reliable solutions for quality monitoring and defect detection in more industries.

[0034] This invention is not limited to the products described above. All technical solutions derived using equivalent substitutions fall within the scope of protection claimed by this invention.

Claims

1. A method for detecting surface tearing on a belt conveyor based on image processing, characterized in that, Includes the following steps: Step 1: Set up a laser and a camera above the belt and then capture images. The laser is directed towards the belt and the laser beam can cover the width of the belt. The camera can capture images of the areas on the belt that are illuminated by the laser. Step 2: Input a reference image of a belt without defects, perform grayscale thresholding on the reference image, and extract the regions where the grayscale value exceeds the grayscale threshold; Step 3: Use morphological dilation to connect adjacent regions, generating a ROI region larger than the actual laser projection line; Step 4: Perform connected component analysis to identify the largest connected region and generate a Boolean mask; Step 5: Acquire the image to be detected captured by the camera, and extract the ROI region image using a Boolean mask; Step 6: Calculate the dynamic threshold of each column in the ROI region image and generate a binary image; Step 7: Divide the binary image into several columns and calculate the cumulative pixel sum for each group; Step 8: Calculate the difference between the cumulative sum of pixels in each group. The position where the difference is greater than 0 indicates that there is a defect on the belt corresponding to that group of pixels.

2. The method for detecting surface tearing of a belt conveyor based on image processing according to claim 1, characterized in that: The specific method of step 6 is to calculate the pixel value of each column in the ROI region image, calculate the average value of pixels whose normalized pixel gray value is less than the gray value threshold, and add an offset as the dynamic threshold of the corresponding column. The pixel values ​​in this column are then compared with the calculated dynamic threshold to generate a binarized result, resulting in a binary image.

3. The method for detecting surface tearing of a belt conveyor based on image processing according to claim 2, characterized in that: The grayscale threshold is 0.9 and the offset is 0.

2.

4. The method for detecting surface tearing of a belt conveyor based on image processing according to claim 1, characterized in that: In step 5, after acquiring the image to be detected captured by the camera, median filtering is first performed.

5. The method for detecting surface tearing of a belt conveyor based on image processing according to claim 1, characterized in that: In step 7, median filtering is performed before grouping the binary images.

6. The method for detecting surface tearing of a belt conveyor based on image processing according to claim 1, characterized in that: In step 2, the grayscale threshold is 90% of the highest value.

7. The method for detecting surface tearing of a belt conveyor based on image processing according to claim 1, characterized in that: The laser emitted by the laser is elongated.

Images