Visual detection system for longitudinal tearing of conveying belt

Abstract

The invention discloses a visual detection system for longitudinal tearing of a conveying belt, relates to the technical field of image recognition, and optimizes the problems of insufficient anti-interference capability, detection precision, false alarm rate and insufficient system integration and expansibility of the visual detection system for the conveying belt. Dust concentration, muddy water attachment and illumination conditions are monitored in real time, and a window structure and a cleaning device are dynamically adjusted by using a logical judgment algorithm, so that environmental interference is effectively reduced, image acquisition quality is improved, and definition of 3D image data is ensured; a multi-detection mechanism and a secondary confirmation strategy are introduced, a deep learning model is utilized to fuse historical data to construct a time sequence model, a tear prediction result is optimized through an attention mechanism, a tear risk assessment index Tlzs is finally generated, the detection precision is improved, and the false alarm rate is reduced. TCP / IP, Modbus Tcp and webSocket protocols are supported, data formats are converted in a unified mode and displayed visually, and centralized management of multiple devices is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and particularly to a visual detection system for longitudinal tearing of a conveyor belt. Background Art

[0002] The conveyor belt visual detection system originated in the late 20th century. With the development of computer vision and industrial automation technologies, it has gradually evolved from simple optoelectronic sensing detection to high-precision, multi-dimensional intelligent detection. In the early days, the detection of conveyor belts mainly relied on manual inspections or mechanical sensors, which were difficult to meet the requirements of high-speed and high-precision detection. Since the 21st century, with the development of high-resolution industrial cameras, deep learning algorithms, and embedded computing devices, the conveyor belt visual detection system has achieved a leap from two-dimensional to three-dimensional, from static to dynamic, and from single-point detection to full-frame monitoring. In recent years, combined with multi-modal imaging technologies such as infrared, laser, and X-ray, the system has been further improved in terms of accuracy, real-time performance, and intelligence.

[0003] The conveyor belt visual detection system is widely used in fields such as minerals, metallurgy, logistics, food, and automotive manufacturing. However, when applied to the safety detection of conveyor belts in the coal mining and mining industries, the following technical drawbacks often exist:

[0004] Insufficient anti-interference ability:

[0005] In the prior art, the camera of the device is easily affected by environmental factors such as dust and muddy water, resulting in unclear image acquisition and affecting the detection accuracy; the window structure design of the early devices was not reasonable enough, and dust and muddy water were easily directly splashed onto the glass surface, affecting the normal operation of the device.

[0006] Problems of detection accuracy and false alarm rate:

[0007] When the early devices detected longitudinal tearing, false alarms or missed detections might occur. Especially when a single-frame image was abnormal, the lack of a secondary confirmation mechanism led to misjudgments; the AI algorithms of the prior art might not be mature enough to quickly and accurately distinguish torn pictures, especially in complex environments.

[0008] Insufficient system integration and scalability:

[0009] The system integration degree of the early devices was relatively low, and centralized monitoring and management of multiple devices could not be achieved; the output interfaces and protocol supports of the prior art were limited, and it was difficult to seamlessly connect with third-party platforms, restricting the scalability and flexibility of the system. Summary of the Invention

[0010] Aiming at the deficiencies of the prior art, the present invention provides a visual detection system for longitudinal tearing of a conveyor belt, which solves the technical drawbacks mentioned in the background art.

[0011] To achieve the above objectives, the present invention is implemented through the following technical solutions: A visual detection system for longitudinal tearing of a conveyor belt, comprising an image acquisition module, an AI recognition module, an anti-interference optimization module, a remote centralized monitoring module, and a feedback management module;

[0012] The image acquisition module is used to collect 3D image data of the bottom surface of the conveyor belt in real time through a laser and a camera, preprocess the collected 3D image data, and finally construct a 3D image data set and transmit it to the AI recognition module;

[0013] The AI recognition module is used to extract features from the 3D image data set, gradually calculate and evaluate the tearing probability value Tlgl, the tearing length value Tlcd, and the tearing width value Tlkd, then construct a tearing prediction model, and calculate and obtain the conveyor belt tearing risk assessment index Tlzs;

[0014] The anti-interference optimization module is used to adjust the system stability through logical judgment and multiple detection mechanisms; at the same time, through the logical control window structure and the cleaning device, adjust the influence of dust and muddy water on the camera;

[0015] The remote centralized monitoring module is used to transmit the tearing risk assessment index Tlzs and related image data to the remote monitoring platform through protocols including TCP / IP, Modbus Tcp, and webSocket, and display them in a visual form on the user interface, and at the same time support centralized monitoring of multiple devices;

[0016] The feedback management module is used to conduct logical comparison and analysis between the preset tearing risk threshold P and the tearing risk assessment index Tlzs, evaluate the tearing risk status of the conveyor belt in real time, and feedback control instructions including alarm, shutdown, and start of the cleaning device through the user interface.

[0017] Preferably, the image acquisition module includes a laser and a camera. The laser is installed at a preset position on the bottom surface of the conveyor belt and is used to emit strip-shaped laser upward to the bottom surface of the conveyor belt. The camera is installed at a position corresponding to the laser and is used to collect image data of the laser reflected on the bottom surface of the conveyor belt in real time; the laser and the camera work synchronously. The laser stripe emitted by the laser forms continuous scanning on the surface of the conveyor belt. The camera captures the reflected image of the laser stripe at a fixed frame rate, and performs denoising, enhancement, and normalization processing on the captured image data, and extracts the center line of the laser stripe, and generates 3D image data of the bottom surface of the conveyor belt through a three-dimensional reconstruction algorithm; the preprocessed 3D image data is constructed into a 3D image data set according to the time series, and the 3D image data set is transmitted to the AI recognition module in real time through the data transmission interface for subsequent feature extraction and tearing recognition.

[0018] Preferably, the AI recognition module includes a feature extraction unit, a probability evaluation unit, a geometric calculation unit, and a model construction unit;

[0019] The feature extraction unit is used to receive the 3D image dataset from the image acquisition module, analyze each frame of the image in the dataset through an image processing algorithm, and extract the morphological change features of the laser stripe on the bottom surface of the conveyor belt, including the continuity, curvature, and local distortion of the stripe; at the same time, based on the morphological change features, potential crack areas are identified, and the geometric features of the cracks are extracted, including the length, width, and depth information of the cracks; in addition, the texture feature data of the conveyor belt surface is extracted through a texture analysis algorithm, including surface roughness, wear degree, and foreign object attachment conditions; after the extracted texture feature data is standardized, it is output to the probability calculation unit and the geometric calculation unit in a preset format.

[0020] Preferably, the probability evaluation unit extracts the surface roughness Tal, texture direction consistency Tbl, local contrast Tcl, and foreign object attachment density Tdl in the texture feature data based on the texture feature data, and after dimensionless processing, calculates the tear probability value Tlgl through the following formula:

[0021]

[0022] The preset tear probability threshold T is compared and evaluated with the tear probability value Tlgl, and the specific content is as follows:

[0023] If the tear probability value Tlgl < the tear probability threshold T, it means that the current conveyor belt has no tear risk, the system does not perform further calculations, and continues to monitor the next frame of image data; at the same time, the current state is recorded as "normal", and the historical data is updated;

[0024] If the tear probability value Tlgl ≥ the tear probability threshold T, it means that the current conveyor belt has a tear risk, and the system enters the detailed analysis stage, including further calculating the tear length value Tlcd and the tear width value Tlkd to quantify the severity of the tear; at the same time, the current state is recorded as "abnormal", and the historical data is updated.

[0025] Preferably, the geometric calculation unit calculates the tear length value Tlcd and the tear width value Tlkd respectively based on the texture feature data;

[0026] After extracting the longitudinal crack extension Tac, longitudinal texture fracture Tbc, and longitudinal gray value change Tcc in the texture feature data and performing dimensionless processing, the tear length value Tlcd is calculated, and the specific calculation formula is as follows:

[0027]

[0028] After extracting the transverse crack extension Tad, transverse texture fracture degree Tbd, and transverse grayscale change value Tcd in the texture feature data and performing dimensionless processing, the tearing width value Tlkd is calculated. The specific calculation formula is as follows:

[0029]

[0030] Preferably, the model construction unit is used to construct a tearing prediction model by combining the tearing probability value Tlgl, the tearing length value Tlcd, and the tearing width value Tlkd. First, it receives the tearing probability value Tlgl from the probability calculation unit, the tearing length value Tlcd, and the tearing width value Tlkd from the geometric calculation unit, and inputs the three into a pre-trained deep learning framework. Secondly, through a multi-layer neural network, feature fusion is performed on the tearing probability value Tlgl, the tearing length value Tlcd, and the tearing width value Tlkd, high-dimensional feature vectors are extracted, and a time series model is constructed in combination with historical tearing data to capture the dynamic trend of tearing development. Then, an attention mechanism is introduced to perform weighted processing on the tearing length value Tlcd and the tearing width value Tlkd. Finally, the tearing risk assessment index Tlzs is output through the Softmax function and transmitted to the risk assessment unit.

[0031] Preferably, the anti-interference optimization module includes a logic control unit and a stability optimization unit

[0032] The logic control unit uses an environmental sensor to continuously monitor the environmental interference factors around the camera, including dust concentration, muddy water adhesion, and lighting conditions; analyzes the environmental interference data using a logic judgment algorithm, and dynamically adjusts the window structure and the working mode of the cleaning device; then triggers the cleaning device to perform automatic cleaning, and balances the impact of interference objects on the camera by adjusting the window height and angle. At the same time, when the lighting condition is insufficient, the exposure parameters of the camera are automatically adjusted.

[0033] Preferably, the stability optimization unit includes a multiple detection mechanism that analyzes a single-frame image multiple times through the multiple detection mechanism; at the same time, secondary confirmation is performed on abnormal images, and an alarm is triggered only when it is continuously judged as torn multiple times; then, using a stability optimization algorithm, the running state of the system is monitored in real time, the system stability is balanced through a logic control algorithm, and in combination with the adjustment results of the environmental interference detection and the logic control unit, the anti-interference ability of the system is further adjusted.

[0034] Preferably, the remote centralized monitoring module is used to receive the tearing risk assessment index Tlzs and related image data, and transmit the data to the remote monitoring platform through TCP / IP, Modbus Tcp and webSocket protocols; then uniformly convert the data of different protocols into a standard format, and then display the received data in the form of charts, images and texts on the user interface, real-time display the running status, tearing risk assessment index Tlzs and alarm information of each conveyor belt, and at the same time support the centralized monitoring and management of multiple devices; the user can view the running status of each conveyor belt in real time through the user interface and take corresponding control instructions according to the feedback information.

[0035] Preferably, the feedback management module is used to preset the tearing risk threshold P according to historical tearing data, conveyor belt running environment and equipment performance; then conduct real-time comparative analysis between the tearing risk assessment index Tlzs and the tearing risk threshold P, and the specific assessment content is as follows:

[0036] When the tearing risk assessment index Tlzs ≥ the tearing risk threshold P, it is determined that the conveyor belt has a tearing risk;

[0037] When the tearing risk assessment index Tlzs < the tearing risk threshold P, it is determined that the conveyor belt is running normally;

[0038] Finally, corresponding control instructions are generated according to the evaluation content of the tearing risk assessment index Tlzs, including triggering an alarm, stopping the machine or starting the cleaning device, and feedback the instructions to the operator through the user interface.

[0039] The present invention provides a visual detection system for longitudinal tearing of a conveyor belt. It has the following beneficial effects:

[0040] (1) For this visual detection system for longitudinal tearing of a conveyor belt, aiming at solving the problem of insufficient anti-interference ability: in the prior art, the camera of the conveyor belt visual detection system is easily affected by environmental factors such as dust and muddy water, resulting in unclear image acquisition and affecting the accuracy of tearing detection; in view of this problem, through the coordinated action of the logic control unit and the stability optimization unit of the anti-interference optimization module of the present invention, the dust concentration, muddy water adhesion and light conditions around the camera are monitored in real time, and the window structure and the working mode of the cleaning device are dynamically adjusted by using the logical judgment algorithm; by logically controlling the window structure and the cleaning device, the environmental interference is effectively reduced, the image acquisition quality is improved, and the clarity of the 3D image data of the bottom surface of the conveyor belt is ensured, thereby improving the stability and reliability of tearing detection.

[0041] (2) For this longitudinal tear visual detection system of a conveyor belt, the traditional conveyor belt tear detection system may have false alarms or missed alarms when detecting longitudinal tears. Especially when a single-frame image is abnormal, there is a lack of a secondary confirmation mechanism, resulting in misjudgment of the system. Therefore, in the AI recognition module of the present invention, a multiple detection mechanism and a secondary confirmation strategy are introduced. The probability evaluation unit first calculates the tear probability value Tlgl based on the texture feature data, and combines the longitudinal crack extension Tac, the longitudinal texture breakage degree Tbc, and the longitudinal gray-scale change value Tcc to calculate the tear length value Tlcd. At the same time, the lateral crack expansion Tad, the lateral texture breakage degree Tbd, and the lateral gray-scale change value Tcd are extracted to calculate the tear width value Tlkd. On this basis, the model construction unit performs feature fusion on Tlgl, Tlcd, and Tlkd through a multi-layer neural network, and constructs a time series model in combination with historical tear data to capture the dynamic trend of tear development. The attention mechanism is introduced to perform weighted processing on the tear length value Tlcd and the tear width value Tlkd, and the tear risk assessment index Tlzs is output through the Softmax function, so as to achieve high-precision analysis of tear detection and effectively reduce the false alarm rate.

[0042] (3) For this longitudinal tear visual detection system of a conveyor belt, the integration of the traditional conveyor belt visual detection system is relatively low, making it difficult to achieve remote centralized monitoring and multi-device management. At the same time, there are bottlenecks in data transmission and system expansion. To address this issue, the remote centralized monitoring module of the present invention supports TCP / IP, Modbus Tcp, and webSocket protocols, transmits the tear risk assessment index Tlzs and related image data to the remote monitoring platform, and converts them into a standard format for visual display to achieve centralized management of multiple devices. The feedback management module sets the tear risk threshold P based on historical tear data, the conveyor belt operating environment, and equipment performance, and compares and evaluates the relationship between Tlzs and P in real time. When Tlzs≥P, the system determines that there is a tear risk in the conveyor belt, automatically triggers an alarm, shuts down, or starts the cleaning device, and feeds back to the operator through the user interface to ensure the stable operation of the system. At the same time, a standardized data interface is provided to enhance the compatibility and expansion ability of the system. Description of the Drawings

[0043] Figure 1 It is a schematic diagram of the framework structure of a longitudinal tear visual detection system for a conveyor belt according to the present invention. Detailed Embodiments

[0044] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0045] Embodiment 1

[0046] Please refer to Figure 1 , the present invention provides a visual detection system for longitudinal tearing of a conveyor belt, including an image acquisition module, an AI recognition module, an anti-interference optimization module, a remote centralized monitoring module, and a feedback management module;

[0047] The image acquisition module is used to collect 3D image data of the bottom surface of the conveyor belt in real time through a laser and a camera, preprocess the collected 3D image data, and finally construct a 3D image data set and transmit it to the AI recognition module;

[0048] The AI recognition module is used to extract features from the 3D image data set, gradually calculate and evaluate the tearing probability value Tlgl, the tearing length value Tlcd, and the tearing width value Tlkd, then construct a tearing prediction model, and calculate and obtain the conveyor belt tearing risk assessment index Tlzs;

[0049] The anti-interference optimization module is used to adjust the system stability through logical judgment and multiple detection mechanisms; at the same time, through the logical control window structure and the cleaning device, adjust the influence of dust and muddy water on the camera;

[0050] The remote centralized monitoring module is used to transmit the tearing risk assessment index Tlzs and related image data to the remote monitoring platform through protocols including TCP / IP, Modbus Tcp, and webSocket, and display them in a visual form on the user interface, and at the same time support the centralized monitoring of multiple devices;

[0051] The feedback management module is used to perform logical comparison and analysis with the tearing risk assessment index Tlzs according to the preset tearing risk threshold P, real-time evaluate the tearing risk status of the conveyor belt, and feedback control instructions including alarm, shutdown, and start the cleaning device through the user interface.

[0052] In this embodiment, the image acquisition module can collect 3D image data of the bottom surface of the conveyor belt in real time through a laser and a camera. After preprocessing the image by denoising, enhancing, and normalizing, a 3D image data set is constructed and accurately transmitted to the AI recognition module, thus ensuring data integrity and real-time performance; the AI recognition module uses image processing algorithms to extract features from the 3D image data set, gradually calculates the tear probability value Tlgl, the tear length value Tlcd, and the tear width value Tlkd, and constructs a tear prediction model to calculate and obtain the conveyor belt tear risk assessment index Tlzs, realizing the precise quantification and prediction of the tear risk; the anti-interference optimization module adjusts the system stability in real time through logical judgment and multiple detection mechanisms. At the same time, using the logical control window structure and the cleaning device, it effectively reduces the interference of dust and muddy water on the camera and ensures the clarity of image acquisition; the remote centralized monitoring module uses protocols such as TCP / IP, Modbus Tcp, and webSocket to transmit the tear risk assessment index Tlzs and related image data to the remote monitoring platform and displays them in a visual form on the user interface, realizing the centralized monitoring and management of multiple devices; the feedback management module performs real-time logical comparison and analysis based on the preset tear risk threshold P and Tlzs, and timely feedbacks control instructions such as alarm, shutdown, and start of the cleaning device, thus realizing the effective early warning and control of the conveyor belt tear risk.

[0053] Embodiment 2

[0054] The image acquisition module includes a laser and a camera. The laser is installed at a preset position on the bottom surface of the conveyor belt and is used to emit a strip of laser upward to the bottom surface of the conveyor belt. The camera is installed at a position corresponding to the laser and is used to collect the image data of the laser reflected on the bottom surface of the conveyor belt in real time; the laser and the camera work synchronously. The laser stripe emitted by the laser forms a continuous scan on the surface of the conveyor belt. The camera captures the reflected image of the laser stripe at a fixed frame rate, and performs denoising, enhancing, and normalizing processing on the captured image data, and extracts the center line of the laser stripe. The 3D image data of the bottom surface of the conveyor belt is generated through a three-dimensional reconstruction algorithm; the preprocessed 3D image data is constructed into a 3D image data set according to the time series, and the 3D image data set is transmitted to the AI recognition module in real time through the data transmission interface for subsequent feature extraction and tear recognition.

[0055] The AI recognition module includes a feature extraction unit, a probability evaluation unit, a geometric calculation unit, and a model construction unit;

[0056] The feature extraction unit is used to receive the 3D image dataset from the image acquisition module, analyze each frame of the image in the dataset through image processing algorithms, and extract the morphological change features of the laser stripe on the bottom surface of the conveyor belt, including the continuity, curvature, and local distortion of the stripe; at the same time, identify potential crack areas based on the morphological change features and extract the geometric features of the cracks, including the length, width, and depth information of the cracks; in addition, extract the texture feature data of the conveyor belt surface through texture analysis algorithms, including surface roughness, wear degree, and foreign object attachment conditions; after the extracted texture feature data is standardized, it is output to the probability calculation unit and the geometric calculation unit in a preset format.

[0057] The probability evaluation unit extracts the surface roughness Tal, texture direction consistency Tbl, local contrast Tcl, and foreign object attachment density Tdl in the texture feature data based on the texture feature data. After dimensionless processing, the tearing probability value Tlgl is calculated through the following formula:

[0058]

[0059] The preset tearing probability threshold T is compared and evaluated with the tearing probability value Tlgl. The specific content is as follows:

[0060] If the tearing probability value Tlgl < the tearing probability threshold T, it means that there is no tearing risk for the current conveyor belt, and the system does not perform further calculations and continues to monitor the next frame of image data; at the same time, record the current state as "normal" and update the historical data;

[0061] If the tearing probability value Tlgl ≥ the tearing probability threshold T, it means that there is a tearing risk for the current conveyor belt, and the system enters the detailed analysis stage, including further calculating the tearing length value Tlcd and the tearing width value Tlkd to quantify the severity of the tear; at the same time, record the current state as "abnormal" and update the historical data.

[0062] The geometric calculation unit calculates the tearing length value Tlcd and the tearing width value Tlkd respectively based on the texture feature data;

[0063] Extract the longitudinal crack extension Tac, longitudinal texture fracture degree Tbc, and longitudinal gray value change Tcc in the texture feature data. After dimensionless processing, the tearing length value Tlcd is calculated. The specific calculation formula is as follows:

[0064]

[0065] Extract the transverse crack expansion Tad, transverse texture fracture degree Tbd, and transverse gray value change Tcd in the texture feature data. After dimensionless processing, the tearing width value Tlkd is calculated. The specific calculation formula is as follows:

[0066]

[0067] The model construction unit is used to construct a tear prediction model by combining the tear probability value Tlgl, the tear length value Tlcd, and the tear width value Tlkd. First, it receives the tear probability value Tlgl from the probability calculation unit, as well as the tear length value Tlcd and the tear width value Tlkd from the geometric calculation unit, and inputs the three into a pre-trained deep learning framework. Secondly, it performs feature fusion on the tear probability value Tlgl, the tear length value Tlcd, and the tear width value Tlkd through a multi-layer neural network, extracts high-dimensional feature vectors, and constructs a time series model in combination with historical tear data to capture the dynamic trend of tear development. Then, it introduces an attention mechanism to perform weighted processing on the tear length value Tlcd and the tear width value Tlkd. Finally, it outputs the tear risk assessment index Tlzs through the Softmax function and transmits it to the risk assessment unit.

[0068] In this embodiment, through the collaborative action of the image acquisition module, the AI recognition module and its internal units, precise detection and evaluation of longitudinal tearing of the conveyor belt are achieved; the image acquisition module uses a laser and a camera to work synchronously to ensure continuous scanning on the bottom surface of the conveyor belt, obtain high-precision 3D image data, and improve the image quality through denoising, enhancement and normalization processing to provide reliable data for subsequent AI recognition; the feature extraction unit in the AI recognition module deeply analyzes the 3D image data set, extracts the morphological change features of the laser stripes to identify the crack area, calculates the length, width and depth of the crack, and at the same time extracts texture features such as surface roughness Tal, wear degree, texture direction consistency Tbl, local contrast Tcl and foreign object adhesion density Tdl to ensure the comprehensiveness of crack detection; the probability evaluation unit calculates the tearing probability value Tlgl using the texture feature data and compares it with the tearing probability threshold T to determine whether there is a tearing risk for the current conveyor belt. When Tlgl≥T, the system enters the geometric analysis stage to quantify the severity of the tear; the geometric calculation unit calculates the tearing length value Tlcd based on the longitudinal crack extension Tac, the longitudinal texture fracture degree Tbc and the longitudinal gray value change Tcc, and at the same time calculates the tearing width value Tlkd based on the transverse crack expansion Tad, the transverse texture fracture degree Tbd and the transverse gray value change Tcd, so as to accurately measure the spatial size and morphological characteristics of the crack; the model construction unit constructs a deep learning model by combining the tearing probability value Tlgl, the tearing length value Tlcd and the tearing width value Tlkd, uses time series analysis to capture the development trend of the crack, and optimizes the tearing prediction through the attention mechanism to make the calculation results more accurate. Finally, the tearing risk assessment index Tlzs is output through the Softmax function, providing a reliable basis for remote monitoring and early warning; the multi-level analysis mechanism of this system ensures the high precision and stability of detection, making it possible to identify, accurately quantify and predict the trend of conveyor belt tearing at an early stage, thus effectively improving the safety and operation efficiency of the conveying system;

[0069] In addition, the specific calculation formula of the tearing risk assessment index Tlzs is as follows:

[0070] Tlzs = w1×Tlgl + w2×Tlcd + w3×Tlkd

[0071] Among them, the weight coefficients w1, w2, and w3 are dynamically calculated through the attention mechanism to ensure that the model can adaptively adjust the weights according to the importance of the input parameters, so as to more accurately evaluate the tearing risk of the conveyor belt.

[0072] Embodiment 3

[0073] The anti-interference optimization module includes a logic control unit and a stability optimization unit

[0074] The logic control unit monitors the environmental interference factors around the camera in real time through environmental sensors, including dust concentration, muddy water adhesion, and lighting conditions; analyzes the environmental interference data using logical judgment algorithms, and dynamically adjusts the window structure and the working mode of the cleaning device; then triggers the cleaning device to perform automatic cleaning, and balances the impact of interference on the camera by adjusting the window height and angle. At the same time, it automatically adjusts the exposure parameters of the camera when the lighting conditions are insufficient.

[0075] The stability optimization unit includes multiple detection mechanisms, which analyze a single-frame image multiple times through the multiple detection mechanisms; at the same time, it performs secondary confirmation on abnormal images and triggers an alarm only when it is continuously judged as torn multiple times; then uses the stability optimization algorithm to monitor the running state of the system in real time, balances the system stability through logical control algorithms, and further adjusts the anti-interference ability of the system in combination with the adjustment results of the environmental interference detection and the logic control unit.

[0076] The remote centralized monitoring module is used to receive the tear risk assessment index Tlzs and related image data, and transmit the data to the remote monitoring platform through TCP / IP, Modbus Tcp, and webSocket protocols; then unifies the conversion of data in different protocols into a standard format, and then displays the received data in the form of charts, images, and texts on the user interface, and displays the running state, tear risk assessment index Tlzs, and alarm information of each conveyor belt in real time. At the same time, it supports the centralized monitoring and management of multiple devices; users can view the running status of each conveyor belt in real time through the user interface and take corresponding control instructions according to the feedback information.

[0077] The feedback management module is used to preset the tear risk threshold P according to historical tear data, conveyor belt operating environment, and equipment performance; then perform real-time comparative analysis of the tear risk assessment index Tlzs and the tear risk threshold P. The specific assessment content is as follows:

[0078] When the tear risk assessment index Tlzs ≥ the tear risk threshold P, it is determined that there is a tear risk on the conveyor belt;

[0079] When the tear risk assessment index Tlzs < the tear risk threshold P, it is determined that the conveyor belt is operating normally;

[0080] Finally, corresponding control instructions are generated according to the assessment content of the tear risk assessment index Tlzs, including triggering an alarm, stopping the machine, or starting the cleaning device, and the instructions are fed back to the operator through the user interface.

[0081] In this embodiment, the anti-interference optimization module plays a role in improving the detection stability and environmental adaptability in the visual detection system for longitudinal tearing of the conveyor belt; the logic control unit monitors the dust concentration, mud adhesion, and lighting conditions around the camera in real time through environmental sensors, analyzes the environmental interference data using logical judgment algorithms, and dynamically adjusts the window structure and the working mode of the cleaning device to keep the camera always in a clear field of view; when dust and mud are detected to affect the camera, the system automatically triggers the cleaning device to clean, reduces the occlusion of the interfering objects on the camera by adjusting the window height and angle, and dynamically adjusts the exposure parameters of the camera when the lighting conditions are insufficient to ensure the stability and reliability of the collected 3D image data, thereby enhancing the environmental adaptability of the system; the stability optimization unit adopts a multiple detection mechanism, analyzes a single-frame image multiple times, and performs secondary confirmation on abnormal images, and only triggers an alarm when tearing is continuously detected multiple times to avoid false alarms and missed alarms; in addition, the stability optimization algorithm can monitor the running state of the system in real time, and combine with the environmental interference detection results to balance the system stability through logical control algorithms, further improving the anti-interference ability of the system and making the detection results more credible;

[0082] The remote centralized monitoring module is responsible for the remote transmission of data, protocol conversion, and multi-device visual monitoring in the system, realizing the intelligent management of the conveyor belt tearing risk; the module receives the tearing risk assessment index Tlzs and related image data, and transmits the data to the remote monitoring platform through TCP / IP, Modbus Tcp, and webSocket protocols to ensure the stable transmission of data under different network protocols; then, the system performs standardized conversion on the data of different protocols and displays them in the form of charts, images, and texts on the user interface, and the running state, tearing risk assessment index Tlzs, and alarm information of each conveyor belt can be viewed in real time to achieve centralized management; the user can remotely view the running status of the conveyor belt through the user interface and take corresponding control instructions based on the risk information to ensure the remote monitoring and rapid response of the conveyor belt tearing detection and improve the system operation and maintenance efficiency;

[0083] The feedback management module in the system is responsible for the intelligent assessment of tearing risk and the execution of control strategies. By using historical tearing data, the operating environment of the conveyor belt, and equipment performance, it sets the tearing risk threshold P and conducts real-time comparative analysis on the tearing risk assessment index Tlzs to ensure that the system can accurately judge the tearing state of the conveyor belt. When the tearing risk assessment index Tlzs ≥ the tearing risk threshold P, the system determines that there is a tearing risk in the conveyor belt and immediately triggers control instructions, including alarm, shutdown, or starting the cleaning device, to prevent the tearing from deteriorating or further damaging the conveyor belt. When the tearing risk assessment index Tlzs < the tearing risk threshold P, the system determines that the conveyor belt is operating normally and continues to monitor. Finally, the system automatically generates control instructions based on the assessment content of Tlzs and feeds them back to the operator through the user interface to ensure a rapid response in case of tearing and improve the safety of the conveyor belt and the intelligent level of equipment management.

[0084] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A conveyor belt longitudinal tear visual detection system, characterized by: Including image acquisition module, AI recognition module, anti-interference optimization module, remote centralized monitoring module and feedback management module; The image acquisition module is used to collect 3D image data of the bottom surface of the conveyor belt in real time through lasers and cameras, and pre-process the collected 3D image data, and finally construct a 3D image data set and transmit it to the AI ​​recognition module; The AI ​​recognition module is used to extract features from the 3D image data set, and gradually calculate and evaluate the tear probability value Tlgl, tear length value Tlcd and tear width value Tlkd, and then build a tear prediction model to calculate and obtain the conveyor belt tear risk assessment index Tlzs; The anti-interference optimization module is used to adjust the system stability through logical judgment and multiple detection mechanisms; At the same time, the window structure and cleaning device are logically controlled to adjust the impact of dust and mud on the camera; The remote centralized monitoring module is used to transmit the tear risk assessment index Tlzs and related image data to the remote monitoring platform through protocols including TCP / IP, ModbusTcp and webSocket, and display them in a visual form on the user interface, while supporting centralized monitoring of multiple devices; The feedback management module is used to perform logical comparison and analysis based on the preset tear risk threshold P and the tear risk assessment index Tlzs, to evaluate the tear risk status of the conveyor belt in real time, and to feedback control instructions including alarm, shutdown and starting of cleaning device through the user interface.

2. A conveyor belt longitudinal tear visual detection system according to claim 1, characterized in that: The image acquisition module includes a laser and a camera. The laser is installed at a preset position on the bottom of the conveyor belt and is used to emit a strip of laser upward to the bottom of the conveyor belt. The camera is installed at a position corresponding to the laser and is used to collect image data after the laser is reflected from the bottom of the conveyor belt in real time. The laser and the camera work synchronously. The laser stripes emitted by the laser form a continuous scan on the surface of the conveyor belt. The camera captures the reflected image of the laser stripes at a fixed frame rate, denoises, enhances and standardizes the captured image data, extracts the center line of the laser stripes, and generates 3D image data of the bottom of the conveyor belt through a three-dimensional reconstruction algorithm. The preprocessed 3D image data is constructed into a 3D image data set in time series, and the 3D image data set is transmitted to the AI ​​recognition module in real time through the data transmission interface for subsequent feature extraction and tear recognition.

3. A conveyor belt longitudinal tear visual detection system method according to claim 2, characterized in that: The AI ​​recognition module includes a feature extraction unit, a probability evaluation unit, a geometric calculation unit, and a model building unit; The feature extraction unit is used to receive the 3D image data set from the image acquisition module, analyze each frame of the data set through the image processing algorithm, and extract the morphological change characteristics of the laser stripes on the bottom surface of the conveyor belt, including the continuity, curvature and local distortion of the stripes; at the same time, identify potential crack areas based on the morphological change characteristics, and extract the geometric characteristics of the cracks, including the length, width and depth information of the cracks; in addition, extract the texture feature data of the conveyor belt surface through the texture analysis algorithm, including the surface roughness, degree of wear and foreign matter adhesion; the extracted texture feature data is standardized and output to the probability calculation unit and the geometric calculation unit in a preset format.

4. A conveyor belt longitudinal tear visual detection system according to claim 3, characterized in that: The probability evaluation unit extracts the surface roughness Tal, texture direction consistency Tbl, local contrast Tcl, and foreign matter attachment density Tdl from the texture feature data based on the texture feature data, and after dimensionless processing, calculates the tearing probability value Tlgl through the following formula: The tearing probability threshold T is preset and compared with the tearing probability value Tlgl for evaluation. The specific contents are as follows: If the tear probability value Tlgl is less than the tear probability threshold T, it means that there is no tear risk on the current conveyor belt. The system does not perform further calculations and continues to monitor the next frame of image data. At the same time, the current state is recorded as "normal" and the historical data is updated. If the tear probability value Tlgl ≥ tear probability threshold T, it means that there is a risk of tearing in the current conveyor belt, and the system enters the detailed analysis stage, including further calculating the tear length value Tlcd and the tear width value Tlkd to quantify the severity of the tear; at the same time, the current status is recorded as "abnormal" and the historical data is updated.

5. A conveyor belt longitudinal tear visual detection system according to claim 4, characterized in that: The geometric calculation unit calculates the tear length value Tlcd and the tear width value Tlkd based on the texture feature data; After extracting the longitudinal crack extension Tac, longitudinal texture fracture Tbc and longitudinal grayscale change value Tcc from the texture feature data and performing dimensionless processing, the tear length value Tlcd is calculated. The specific calculation formula is as follows: After extracting the transverse crack extension Tad, transverse texture fracture degree Tbd and transverse grayscale change value Tcd from the texture feature data and performing dimensionless processing, the tear width value Tlkd is calculated. The specific calculation formula is as follows:

6. A conveyor belt longitudinal tear visual detection system according to claim 5, characterized in that: The model building unit is used to combine the tear probability value Tlgl, the tear length value Tlcd and the tear width value Tlkd to build a tear prediction model; first, the tear probability value Tlgl from the probability calculation unit and the tear length value Tlcd and the tear width value Tlkd from the geometry calculation unit are received, and the three are input into the pre-trained deep learning framework; secondly, the tear probability value Tlgl, the tear length value Tlcd and the tear width value Tlkd are feature fused through a multi-layer neural network to extract high-dimensional feature vectors, and a time series model is built in combination with historical tear data to capture the dynamic trend of tear development; then, an attention mechanism is introduced to perform weighted processing on the tear length value Tlcd and the tear width value Tlkd; finally, the tear risk assessment index Tlzs is output through the Softmax function and transmitted to the risk assessment unit.

7. A conveyor belt longitudinal tear visual detection system according to claim 6, characterized in that: The anti-interference optimization module includes a logic control unit and a stability optimization unit. The logic control unit uses environmental sensors to monitor the environmental interference factors around the camera in real time, including dust concentration, mud and water adhesion, and lighting conditions. It uses logic judgment algorithms to analyze environmental interference data and dynamically adjust the window structure and the working mode of the cleaning device. It then triggers the cleaning device to perform automatic cleaning, and balances the impact of interference on the camera by adjusting the window height and angle. It also automatically adjusts the camera's exposure parameters when lighting conditions are insufficient.

8. A conveyor belt longitudinal tear visual detection system according to claim 7, characterized in that: The stability optimization unit includes multiple detection mechanisms, through which a single-frame image is analyzed multiple times. At the same time, abnormal images are reconfirmed and an alarm is triggered only when they are judged to be torn for multiple consecutive times. The stability optimization algorithm is then used to monitor the system's operating status in real time, and the system stability is balanced through the logic control algorithm. The system's anti-interference ability is further adjusted by combining the environmental interference detection and the adjustment results of the logic control unit.

9. A conveyor belt longitudinal tear visual detection system according to claim 8, characterized in that: The remote centralized monitoring module is used to receive the tear risk assessment index Tlzs and related image data, and transmit the data to the remote monitoring platform through TCP / IP, Modbus Tcp and webSocket protocols; then the data of different protocols are uniformly converted into a standard format, and then the received data is displayed on the user interface in the form of charts, images and texts, and the operating status, tear risk assessment index Tlzs and alarm information of each conveyor belt are displayed in real time, and centralized monitoring and management of multiple devices are supported at the same time; users can view the operating status of each conveyor belt in real time through the user interface, and take corresponding control instructions according to the feedback information.

10. A conveyor belt longitudinal tear visual detection system according to claim 9, characterized in that: The feedback management module is used to preset the tear risk threshold P according to the historical tear data, conveyor belt operating environment and equipment performance; then the tear risk assessment index Tlzs is compared and analyzed with the tear risk threshold P in real time. The specific evaluation contents are as follows: When the tear risk assessment index Tlzs ≥ the tear risk threshold P, it is determined that the conveyor belt has a tear risk; When the tear risk assessment index Tlzs < the tear risk threshold P, the conveyor belt is judged to be operating normally; Finally, the corresponding control instructions are generated according to the evaluation content of the tear risk assessment index Tlzs, including triggering an alarm, shutting down or starting a cleaning device, and the instructions are fed back to the operator through the user interface.

Images