Buckle type adhesive tape joint damage detection method and system based on machine vision
By using machine vision inspection methods and employing YOLOv11 and Deepsort algorithms to detect tape joint damage, this method solves the problems of low efficiency and insufficient accuracy of traditional inspection methods, and achieves high-precision and real-time monitoring of tape joint damage.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 云鼎科技股份有限公司
- Filing Date
- 2025-12-03
- Publication Date
- 2026-04-24
AI Technical Summary
Traditional snap-on tape joint detection methods are inefficient and lack precision, failing to meet the requirements of high precision, high efficiency, and real-time performance. Furthermore, they are prone to missed detections or misjudgments due to human factors.
A machine vision-based detection method is adopted, which uses an intrinsically safe camera to acquire images, uses the YOLOv11 algorithm to detect tape joints and gaps, combines the Deepsort algorithm for target tracking, calculates the percentage of the gap to the width of the belt, sets thresholds for minor and severe damage, and provides alarms through voice prompts or control of belt shutdown.
It achieves high-precision, real-time detection of belt joint damage, reduces missed detections and false judgments, can respond and control belt operation in a timely manner, and improves the level of automation in detection.
Smart Images
Figure CN121921250A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of machine vision inspection technology, and in particular relates to a method and system for detecting damage to snap-on tape joints based on machine vision. Background Technology
[0002] In coal mine transportation systems, snap-on belt joints are critical components connecting conveyor belts. Damage to these joints can lead to serious problems such as belt breakage and coal leakage, thereby affecting production efficiency and potentially causing safety accidents. For example, an accident occurred in a coal conveying plant where coal accumulation caused malfunctions in the belt tensioning device, ultimately resulting in excessive belt wear and breakage. Furthermore, damaged belt joints can also cause belt misalignment and material spillage, further impacting the normal operation of the transportation system.
[0003] Currently, traditional methods for detecting damage to snap-fit conveyor belt joints mainly rely on manual inspection or simple sensors. Manual inspection is not only inefficient but also prone to missed detections or misjudgments due to human error. While simple sensors can achieve a certain degree of automation, their detection range is limited, their accuracy is insufficient, and they cannot monitor the dynamic changes of the conveyor belt joint in real time. These problems make it difficult for traditional detection methods to meet the high-precision, high-efficiency, and real-time requirements of coal mine transportation systems for conveyor belt joint detection. Summary of the Invention
[0004] (a) Purpose of the invention To overcome the above shortcomings, the present invention aims to provide a machine vision-based method and system for detecting damage to snap-on tape joints, thereby solving the aforementioned technical problems.
[0005] (II) Technical Solution To achieve the above objectives, the technical solution provided in this application is as follows: A machine vision-based method for detecting damage to snap-on tape joints includes the following steps: S1 uses an intrinsically safe camera to pre-capture images of snap-fit tape joints and joint notches to create an image dataset; S2 uses images from the image dataset to train a buckle-type tape joint damage tracking model. The model consists of two parts: a detector and a tracker. The detector uses the YOLOv11 algorithm, and the tracker uses the Deepsort algorithm. S3 When the belt conveyor is started and running, the intrinsically safe camera captures images of the belt and inputs the images into the buckle-type belt joint damage tracking model to obtain the detection results of the belt joint and its gaps; Based on the x-coordinate values of the upper left and lower right corners of the belt interface and joint gap detection frame, S4 calculates the percentage of the gap relative to the belt width. It then determines whether the gap damage rate exceeds the minor damage threshold and the severe damage threshold. If the minor damage threshold is exceeded, the relevant image is uploaded to the system platform and the corresponding camera is activated to provide an on-site voice alert. If the severe damage threshold is exceeded, the relevant evidence image is uploaded to the system platform, and an on-site voice alert and belt shutdown are activated.
[0006] Preferably, the intrinsically safe camera is installed directly above the conveyor belt and shines obliquely downwards, so that the conveyor belt occupies 70% or more of the image and the image resolution is not less than 1920*1080. The captured images are divided into images with and without coal when the conveyor belt is running normally.
[0007] Preferably, the image dataset is labeled using the deep learning image labeling software CVAT, with the label type being rectangular boxes and the label categories being tape joints and notches.
[0008] Preferably, the YOLOv11 algorithm mainly includes the C3K2 module, the C2PSA module, and depthwise separable convolution. The C3K2 module splits the feature map and applies 3×3 small kernel convolution to optimize the information flow, thereby enhancing feature extraction capabilities; The C2PSA module is set in the Neck structure and combines a point spatial attention mechanism to dynamically fuse low-level and high-level features, thereby improving the detection accuracy of small targets and occluded targets in complex backgrounds. The depthwise separable convolution significantly reduces the number of parameters and computational cost by separating the calculation of spatial and channel dimensions, while maintaining classification performance.
[0009] Preferably, the Deepsort algorithm flows as follows: First, tape joints and gaps are detected using the YOLOv11 detector. When tape joints and gaps are detected, a detection box for tape joints and gaps is generated. Then, based on the tape joint and notch trajectory frames of the previous frame, the Kalman filter algorithm is applied to predict the trajectory frame of the next moment; Furthermore, the Hungarian algorithm is applied to perform cascade matching and IOU matching on the predicted trajectory boxes and the detection boxes of the current frame. Cascade matching uses the appearance features of the tape joint and the joint notch to calculate the similarity with motion features, while IOU matching uses the trajectory and the area of the target box to calculate the IOU. Finally, Kalman filtering is applied to update the successfully matched trajectory boxes.
[0010] Preferably, the calculation of the percentage of the gap to the belt width in S4 is achieved through the following operation: Assuming a total of n tape joint notches are detected, the coordinates of the top-left corner of the detection frame for each joint can be obtained based on the target detection model. and the coordinates of the bottom right corner The coordinates of the upper left corner of each joint notch detection box and the coordinates of the bottom right corner , The total width of the damaged tape joint is: ) The width of the belt interface is: From Formula 1 and Formula 2, the notch damage rate can be obtained as follows: The defect rate is used to determine whether the threshold for minor damage or severe damage has been exceeded.
[0011] Preferably, in S4, the threshold for minor damage is a damage value exceeding 10% of the belt width, and the threshold for severe damage is a damage value exceeding 20% of the belt width.
[0012] A machine vision-based system for detecting damage to snap-on tape joints includes the following modules: The image acquisition module enables the intrinsically safe camera in the coal mine to capture high-definition images of the coal conveyor belt during operation. The surface of the intrinsically safe camera is made of dust-repellent material. The data processing module is used to process and analyze the acquired images, including target detection and damage assessment, wherein the target detection adopts the YOLOv11 algorithm; The alarm module is used to issue voice alerts or control the belt to stop based on the degree of damage. A communication module is used to transmit detection results and alarm information to the system platform. The communication module includes wireless transmission and wired transmission methods. The wireless transmission methods include 5G and industrial Wi-Fi, and the wired transmission methods include USB 3.0 and Ethernet. Control module: Used to receive instructions from the system platform and control the operation of the alarm module and belt conveyor.
[0013] Preferably, the data processing module includes: The image preprocessing unit is used to preprocess the acquired images, including image resizing and normalization. Damage Assessment Unit. Based on the detection results, the notch damage rate is calculated and classified according to the degree of damage. When the damage rate exceeds the minor damage threshold, relevant images are uploaded to the system platform, and the corresponding camera is activated to provide on-site voice alerts. When the damage rate exceeds the severe damage threshold, relevant evidence images are uploaded to the system platform, triggering on-site voice alerts and controlling the conveyor belt to stop.
[0014] Preferably, the camera of the image acquisition module is installed directly above the belt and shines obliquely downwards, so that the belt occupies more than 70% of the image, the field of view of the image includes the belt, and the image resolution is not less than 1920×1080. The system also includes a data storage module for storing the acquired images and detection results. The data storage module is an SD card or hard drive with a large storage capacity.
[0015] Beneficial effects: This system employs deep learning-based target tracking to detect belt splice damage and provides tiered alarms based on splice damage rates. When splice damage is detected, it sends voice alerts to staff and halts the conveyor belt, achieving truly online, real-time, and precise monitoring and control. Furthermore, this target tracking model can determine whether splices and gaps are the same, effectively eliminating numerous duplicate alarms and demonstrating high robustness. Attached Figure Description
[0016] Figure 1 This is a flowchart of the present invention; Figure 2 This is a diagram showing the detection results of the present invention. Detailed Implementation
[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the following detailed embodiments are described in conjunction with the appendix. Figure 1-2 The present invention will be described in further detail below. It should be understood that these descriptions are merely exemplary and not intended to limit the scope of the invention. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.
[0018] This invention provides a machine vision-based method for detecting damage to snap-on tape joints, comprising the following steps: S1 uses an intrinsically safe camera to pre-capture images of snap-fit tape joints and joint notches to create an image dataset; S2 uses images from the image dataset to train a buckle-type tape joint damage tracking model. The model consists of two parts: a detector and a tracker. The detector uses the YOLOv11 algorithm, and the tracker uses the Deepsort algorithm. S3 When the belt conveyor is started and running, the intrinsically safe camera captures images of the belt and inputs the images into the buckle-type belt joint damage tracking model to obtain the detection results of the belt joint and its gaps; Based on the x-coordinate values of the upper left and lower right corners of the belt interface and joint gap detection frame, S4 calculates the percentage of the gap relative to the belt width. It then determines whether the gap damage rate exceeds the minor damage threshold and the severe damage threshold. If the minor damage threshold is exceeded, the relevant image is uploaded to the system platform and the corresponding camera is activated to provide an on-site voice alert. If the severe damage threshold is exceeded, the relevant evidence image is uploaded to the system platform, and an on-site voice alert and belt shutdown are activated.
[0019] Preferably, the intrinsically safe camera is installed directly above the conveyor belt and shines obliquely downwards, so that the conveyor belt occupies 70% or more of the image, ensuring that the image field of view includes the conveyor belt and avoiding missing key information due to field of view issues. The image resolution is not less than 1920*1080 to ensure image clarity and help to more accurately detect the damage of the conveyor belt joint. The acquired images are divided into images with and without coal when the conveyor belt is running normally.
[0020] Preferably, the image dataset is labeled using the deep learning image labeling software CVAT, with the label type being rectangular boxes and the label categories being tape joints and notches.
[0021] Preferably, the YOLOv11 algorithm mainly includes the C3K2 module, the C2PSA module, and depthwise separable convolution. The C3K2 module splits the feature map and applies 3×3 small kernel convolution to optimize the information flow, thereby enhancing feature extraction capabilities; The C2PSA module is set in the Neck structure and combines a point spatial attention mechanism to dynamically fuse low-level and high-level features, thereby improving the detection accuracy of small targets and occluded targets in complex backgrounds. The depthwise separable convolution significantly reduces the number of parameters and computational cost by separating the calculation of spatial and channel dimensions, while maintaining classification performance.
[0022] Preferably, the Deepsort algorithm flows as follows: First, tape joints and gaps are detected using the YOLOv11 detector. When tape joints and gaps are detected, a detection box for tape joints and gaps is generated. Then, based on the tape joint and notch trajectory frames of the previous frame, the Kalman filter algorithm is applied to predict the trajectory frame of the next moment; Furthermore, the Hungarian algorithm is applied to perform cascade matching and IOU matching on the predicted trajectory boxes and the detection boxes of the current frame. Cascade matching uses the appearance features of the tape joint and the joint notch to calculate the similarity with motion features, while IOU matching uses the trajectory and the area of the target box to calculate the IOU. Finally, Kalman filtering is applied to update the successfully matched trajectory boxes.
[0023] The Deepsort algorithm is used for target tracking, which can effectively remove duplicate alarms and improve the robustness of the system.
[0024] Preferably, the calculation of the percentage of the gap to the belt width in S4 is achieved through the following operation: Assuming a total of n tape joint notches are detected, the coordinates of the top-left corner of the detection frame for each joint can be obtained based on the target detection model. and the coordinates of the bottom right corner The coordinates of the upper left corner of each joint notch detection box and the coordinates of the bottom right corner , The total width of the damaged tape joint is: ) The width of the belt interface is: From Formula 1 and Formula 2, the notch damage rate can be obtained as follows: The defect rate is used to determine whether the threshold for minor damage or severe damage has been exceeded.
[0025] Preferably, in S4, the threshold for minor damage is a damage value exceeding 10% of the belt width, and the threshold for severe damage is a damage value exceeding 20% of the belt width.
[0026] A machine vision-based system for detecting damage to snap-on tape joints includes the following modules: The image acquisition module enables the intrinsically safe camera in the coal mine to capture high-definition images of the coal conveyor belt during operation. The surface of the intrinsically safe camera is made of dust-repellent material, making it suitable for the underground coal mine environment. The data processing module includes an image preprocessing unit and a damage assessment unit, which are used to process and analyze the acquired images, including target detection and damage assessment, wherein the target detection adopts the YOLOv11 algorithm; The alarm module is used to issue voice alerts or control the belt to stop based on the degree of damage, ensuring timely response; The communication module is used to transmit detection results and alarm information to the system platform. The communication module includes wireless transmission and wired transmission. The wireless transmission methods include 5G and industrial Wi-Fi, and the wired transmission methods include USB 3.0 and Ethernet, to ensure the stability and reliability of data transmission. Control module: Used to receive instructions from the system platform and control the operation of the alarm module and belt conveyor.
[0027] Preferably, the data processing module includes: The image preprocessing unit is used to preprocess the acquired images, including image resizing and normalization. Damage Assessment Unit. Based on the detection results, the notch damage rate is calculated and classified according to the degree of damage. When the damage rate exceeds the minor damage threshold, relevant images are uploaded to the system platform, and the corresponding camera is activated to provide on-site voice alerts. When the damage rate exceeds the severe damage threshold, relevant evidence images are uploaded to the system platform, triggering on-site voice alerts and controlling the conveyor belt to stop.
[0028] Preferably, the camera of the image acquisition module is installed directly above the belt and shines obliquely downwards, so that the belt occupies more than 70% of the image, the field of view of the image includes the belt, and the image resolution is not less than 1920×1080. The system also includes a data storage module for storing the acquired images and detection results. The data storage module is an SD card or hard drive with a large storage capacity.
[0029] In practice, firstly, an intrinsically safe camera is installed directly above the coal conveyor belt in the underground coal mine, ensuring it shines obliquely downwards, with the belt occupying more than 70% of the frame and an image resolution of at least 1920×1080. The camera is then connected to a communication module and an alarm module. Next, images of the snap-fit conveyor belt joints and joint gaps are pre-captured using the intrinsically safe camera, categorized as images with and without coal during normal belt operation. CVAT software is used to annotate the images, using rectangular bounding boxes and classifying them as either belt joints or gaps. The annotated image dataset is saved for model training. Then, the annotated image dataset is used to train a snap-fit conveyor belt joint damage tracking model. This model consists of a detector and a tracker. The detector uses the YOLOv11 algorithm, and the tracker uses the Deepsort algorithm. After training, the model is validated and optimized.
[0030] When the belt conveyor starts operating, the intrinsically safe camera begins to acquire images of the belt in real time. These images are transmitted to the data processing module via the communication module. The data processing module preprocesses the images and then inputs them into a trained model to obtain detection results for belt joints and gaps. Based on the detection results, the percentage of the gap relative to the belt width is calculated to assess the degree of damage. If the damage rate exceeds the minor damage threshold (10%), relevant images are uploaded to the system platform, and a voice alert is triggered. If the damage rate exceeds the severe damage threshold (20%), relevant evidence images are uploaded to the system platform, a voice alert is triggered, and the belt is stopped. Finally, the acquired images and detection results are stored in the data storage module for subsequent analysis and traceability. Regular data backups ensure data security, and the detection process and model parameters are optimized based on the analysis results to improve the overall performance of the system.
[0031] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0032] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A machine vision-based method for detecting damage to snap-on tape joints, characterized in that, Includes the following steps: S1 uses an intrinsically safe camera to pre-capture images of snap-fit tape joints and joint notches to create an image dataset; S2 uses images from the image dataset to train a buckle-type tape joint damage tracking model. The model consists of two parts: a detector and a tracker. The detector uses the YOLOv11 algorithm, and the tracker uses the Deepsort algorithm. S3 When the belt conveyor is started and running, the intrinsically safe camera captures images of the belt and inputs the images into the buckle-type belt joint damage tracking model to obtain the detection results of the belt joint and its gaps; S4 calculates the percentage of the gap to the width of the belt based on the x-coordinate values of the upper left and lower right corners of the tape interface and joint gap detection frame. It determines whether the gap damage rate exceeds the threshold for minor damage and the threshold for severe damage. When the minor damage threshold is exceeded, the relevant image is uploaded to the system platform and the corresponding camera is linked to provide on-site voice reminder. When the severe damage threshold is exceeded, relevant evidence images are uploaded to the system platform, triggering on-site voice prompts and controlling the belt conveyor to stop.
2. The method for detecting damage to snap-on tape joints based on machine vision according to claim 1, characterized in that, The intrinsically safe camera is installed directly above the conveyor belt and shines obliquely downwards, so that the conveyor belt occupies 70% or more of the image and the image resolution is not less than 1920*1080. The captured images are divided into images with and without coal when the conveyor belt is running normally.
3. The method for detecting damage to snap-on tape joints based on machine vision according to claim 2, characterized in that, The image dataset was labeled using the deep learning image labeling software CVAT, with the label type being rectangular boxes and the label categories being tape joints and notches.
4. The method for detecting damage to snap-on tape joints based on machine vision according to claim 1, characterized in that, The YOLOv11 algorithm mainly includes the C3K2 module, the C2PSA module, and depthwise separable convolution. The C3K2 module splits the feature map and applies 3×3 small kernel convolution to optimize the information flow, thereby enhancing feature extraction capabilities; The C2PSA module is set in the Neck structure and combines a point spatial attention mechanism to dynamically fuse low-level and high-level features, thereby improving the detection accuracy of small targets and occluded targets in complex backgrounds. The depthwise separable convolution significantly reduces the number of parameters and computational cost by separating the calculation of spatial and channel dimensions, while maintaining classification performance.
5. The method for detecting damage to snap-on tape joints based on machine vision according to claim 1, characterized in that, The Deepsort algorithm process is as follows: First, tape joints and gaps are detected using the YOLOv11 detector. When tape joints and gaps are detected, a detection box for tape joints and gaps is generated. Then, based on the tape joint and notch trajectory frames of the previous frame, the Kalman filter algorithm is applied to predict the trajectory frame of the next moment; Furthermore, the Hungarian algorithm is applied to perform cascade matching and IOU matching on the predicted trajectory boxes and the detection boxes of the current frame. Cascade matching uses the appearance features of the tape joint and the joint notch to combine motion features to calculate similarity, while IOU matching uses the trajectory and the area of the target box to calculate IOU. Finally, Kalman filtering is applied to update the successfully matched trajectory boxes.
6. The method for detecting damage to snap-on tape joints based on machine vision according to claim 1, characterized in that, The percentage of the gap to the belt width in S4 is calculated through the following operation: Assuming a total of n tape joint notches are detected, the coordinates of the top-left corner of the detection frame for each joint can be obtained based on the target detection model. and the coordinates of the bottom right corner The coordinates of the upper left corner of each joint notch detection box and the coordinates of the bottom right corner , The total width of the damaged tape joint is: ) The width of the belt interface is: From Formula 1 and Formula 2, the notch damage rate can be obtained as follows: The defect rate is used to determine whether the threshold for minor damage or severe damage has been exceeded.
7. The method for detecting damage to snap-on tape joints based on machine vision according to claim 1, characterized in that, In S4, the threshold for minor damage is a damage value exceeding 10% of the belt width, and the threshold for severe damage is a damage value exceeding 20% of the belt width.
8. A machine vision-based system for detecting damage to snap-on tape joints, characterized in that, Includes the following modules: The image acquisition module enables the intrinsically safe camera in the coal mine to capture high-definition images of the coal conveyor belt during operation. The surface of the intrinsically safe camera is made of dust-repellent material. The data processing module is used to process and analyze the acquired images, including target detection and damage assessment, wherein the target detection adopts the YOLOv11 algorithm; The alarm module is used to issue voice alerts or control the belt to stop based on the degree of damage. A communication module is used to transmit detection results and alarm information to the system platform. The communication module includes wireless transmission and wired transmission methods. The wireless transmission methods include 5G and industrial Wi-Fi, and the wired transmission methods include USB 3.0 and Ethernet. Control module: Used to receive instructions from the system platform and control the operation of the alarm module and belt conveyor.
9. A machine vision-based buckle-type tape joint damage detection system according to claim 8, characterized in that, The data processing module includes: The image preprocessing unit is used to preprocess the acquired images, including image resizing and normalization. Damage Assessment Unit. Based on the detection results, the notch damage rate is calculated and classified according to the degree of damage. When the damage rate exceeds the minor damage threshold, relevant images are uploaded to the system platform, and the corresponding camera is activated to provide on-site voice alerts. When the damage rate exceeds the severe damage threshold, relevant evidence images are uploaded to the system platform, triggering on-site voice alerts and controlling the conveyor belt to stop.
10. A machine vision-based buckle-type tape joint damage detection system according to claim 8, characterized in that, The camera of the image acquisition module is installed directly above the belt and shines downwards at an angle, so that the belt occupies more than 70% of the image. The image field of view includes the belt, and the image resolution is not less than 1920×1080. The system also includes a data storage module for storing the acquired images and detection results. The data storage module is an SD card or hard drive with a large storage capacity.