Method and device for monitoring edge wire escaping of circle shear

Abstract

The invention provides a circle shear edge wire escaping monitoring method and device. The problems that in the prior art, circle shear edge wire escaping monitoring is low in accuracy and poor in real-time performance are solved. The method specifically comprises the following steps: S1, acquiring image data used for monitoring an edge wire escaping area of the steel disc shear; step S2, preprocessing the image data to form preprocessed image data; s3, constructing a target detection model; s4, inputting the preprocessed image data into the target detection model to form a target detection result; and S5, performing early warning judgment according to an early warning threshold value and the target detection result to form an early warning judgment result, thereby improving the accuracy and the real-time performance of circle shear edge escaping wire monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of target monitoring, and particularly to a method and device for monitoring escaping edge wires of a circular shear. Background Art

[0002] In the steel processing process, the circular shear plays a key role in shearing the edges of steel plates, and its shearing accuracy and stability are directly related to the quality of the steel plates. However, in the actual operation process, the circular shear may have the phenomenon of escaping edge wires during shearing, that is, the sheared edge wires fail to accurately fall into the preset collection area. This problem not only reduces the cleanliness of the production environment, but may also cause potential damage to the equipment and even pose a serious threat to the safety of operators.

[0003] Traditional means of monitoring escaping edge wires mainly rely on manual inspections. This method is not only inefficient, but also difficult to achieve real-time monitoring and cannot respond to the situation of escaping edge wires in a timely manner. With the continuous development of machine vision and intelligent control technologies, some automated monitoring methods have been tried and applied in the steel processing field to improve this situation. For example, the patent CN202110739737.5 proposes an online monitoring method for escaping wires of a circular shear, which can detect the movement of edge wires on the driving side and the operating side inside the bin in real time through video monitoring images, effectively improving the timeliness of escaping wire monitoring. However, although this method has promoted the progress of automated monitoring to a certain extent, there is still room for improvement in the accuracy and efficiency of monitoring the escaping edge wires of the circular shear. In addition, the patent CN202011034097.X established a deviation prediction model for cold rolling pickling loop strip steel by calculating information such as the center line offset and width change of the hot rolling incoming strip steel, providing a new idea for the risk identification of edge blocking faults of the circular shear. Although this method mainly focuses on the risk identification of edge blocking faults, it also reveals the possibility of using the combination of real-time monitoring data and intelligent algorithms to prevent circular shear-related faults. Further, the patent CN202410586485.0 introduces deep learning technology, and realizes real-time unmanned monitoring of the blockage of waste edge wires of strip steel through the combination of camera image acquisition and algorithm monitoring and recognition modules. This invention not only demonstrates the application potential of intelligent supervision in steel production, but also indicates the great value of deep learning technology in improving the accuracy and efficiency of escaping edge wire monitoring.

[0004] Despite this, the current monitoring methods for escaping edge wires of circular shears still face challenges, and there is an urgent need for a more efficient and accurate automated monitoring method to comprehensively ensure the stable operation of the steel processing process and the safety of operators. Summary of the Invention

[0005] The present invention provides a method and device for monitoring escaping edge wires of a circular shear to solve the problems of low accuracy and poor real-time performance in monitoring escaping edge wires of a circular shear in the prior art.

[0006] In the first aspect, the present invention provides a method for monitoring the edge - escaping wire of a circular shear, specifically including the following steps:

[0007] Step S1: Obtain image data for monitoring the edge - escaping wire area of the circular shear;

[0008] Step S2: Pre - process the image data to form pre - processed image data;

[0009] Step S3: Construct a target detection model;

[0010] Step S4: Input the pre - processed image data into the target detection model to form a target detection result; wherein, the target detection result includes the shape of the steel material, the size of the steel material, the moving speed of the steel material, and the blanking result for monitoring the edge - escaping wire area of the circular shear;

[0011] Step S5: Make a warning judgment according to the warning threshold and the target detection result to form a warning judgment result.

[0012] Preferably, in step S1, the image data of the area to be monitored is obtained through a high - definition camera array and a 3D vision sensor.

[0013] More preferably, the installation positions and installation angles of the high - definition camera array and the 3D vision sensor are optimized by a deep - learning algorithm to ensure clear image capture under complex lighting and occlusion conditions.

[0014] Preferably, in step S2, the pre - processing includes noise reduction.

[0015] More preferably, the image data is denoised by a convolutional neural network (CNN).

[0016] More preferably, an adaptive contrast enhancement technique is also used to dynamically adjust the contrast according to the image content, making the boundary between the steel material target and the background clearer.

[0017] Preferably, some of the data in the image data in step S1 is generated by a generative adversarial network (GAN), or some of the data in the pre - processed image data in step S2 is generated by a generative adversarial network.

[0018] Preferably, in step S3, the target detection model includes one or a combination of multiple models among the R - CNN model, the FPN model (Feature Pyramid Network), the YOLO series models, the RetinaNet model, the CenterNet model, the EfficientDet model, and the DETR model.

[0019] More preferably, the object detection model is the YOLO V7 model integrated with an attention mechanism.

[0020] Preferably, in step S4, before inputting the preprocessed image data into the object detection model, the object detection model is trained by inputting labeled sample data into the object detection model.

[0021] More preferably, in combination with a transfer learning strategy, the process of training the object detection model is accelerated, and the adaptability of the object detection model to data is improved. A multi-threaded parallel processing technology is adopted to detect multiple frames of images simultaneously.

[0022] More preferably, a federated learning strategy is adopted to train the object detection model to protect privacy data and improve the performance of the object detection model.

[0023] More preferably, after the object detection model is trained, the object detection model is evaluated by means of cross-validation and the like to ensure the stability and generalization ability of the model.

[0024] Preferably, in step S4, before inputting the preprocessed image data into the object detection model, it further includes

[0025] Step S401: Input the preprocessed image data into a lightweight neural network for further preliminary detection to form a preliminary detection result;

[0026] Step S402: Input the preliminary detection result into the object detection model to form an object detection result.

[0027] More preferably, in step S401, the lightweight neural network includes one of the YOLO-Tiny model, the MobileNet model, and the ShuffleNet model.

[0028] Preferably, in step S5, early warning judgment is made according to the early warning threshold and the object detection result.

[0029] In this application, the "early warning threshold" is used for early warning judgment.

[0030] More preferably, according to the early warning threshold and the object detection result, an early warning judgment is made in combination with a random forest (RF) classification model.

[0031] Preferably, a method for detecting escaping edge wires of a circular shear further includes

[0032] Step S6: According to the early warning judgment result, judge whether early warning is needed and process the blockage or escape of edge wires of the circular shear. If necessary, give an early warning and process the blockage or escape of edge wires of the circular shear.

[0033] Step S7: After processing the blockage wire or the escaping edge wire of the disk shear, adjust the disk shear parameters and / or the warning threshold, and repeat Steps S4 - S6 until there is no abnormality in the disk shear (in this application, the "abnormality" means that it is necessary to process the blockage wire or the escaping edge wire of the disk shear).

[0034] In this application, the types of the warning threshold include the shape of the steel material, the size of the steel material, the speed of the steel material, and time.

[0035] Second aspect, the present invention also provides a monitoring device for the escaping edge wire of a disk shear, which specifically includes the following modules:

[0036] An image data acquisition module, which is used to acquire image data for monitoring the area of the escaping edge wire of the disk shear;

[0037] An image data preprocessing module, which is used to preprocess the image data to form preprocessed image data;

[0038] A model construction module, which is used to construct an object detection model;

[0039] A detection result generation module, which is used to input the preprocessed image data into the object detection model to form an object detection result;

[0040] A judgment result generation module, which is used to perform a warning judgment according to the object detection result to form a warning judgment result.

[0041] Preferably, in the image data acquisition module, the image data of the area to be monitored is acquired through a high-definition camera array and a 3D vision sensor.

[0042] More preferably, the installation positions and installation angles of the high-definition camera array and the 3D vision sensor are optimized through a deep learning algorithm to ensure clear image capture under complex lighting and occlusion conditions.

[0043] Preferably, in the image data preprocessing module, the preprocessing includes noise reduction.

[0044] More preferably, the image data is denoised through a convolutional neural network (CNN).

[0045] More preferably, an adaptive contrast enhancement technique is also used to dynamically adjust the contrast according to the image content, making the boundary between the steel material target and the background clearer.

[0046] Preferably, part of the data in the image data in the image data acquisition module is generated through a generative adversarial network (GAN), or part of the data in the preprocessed image data in the image data preprocessing module is generated through a generative adversarial network.

[0047] Preferably, in the model construction module, the object detection model includes one or a combination of more than one of the R-CNN model, the FPN model (Feature Pyramid Network), the YOLO series of models, the RetinaNet model, the CenterNet model, the EfficientDet model, and the DETR model.

[0048] More preferably, the object detection model is the YOLO V7 model integrated with an attention mechanism.

[0049] Preferably, in the detection result generation module, before inputting the preprocessed image data into the object detection model, the object detection model is trained by inputting labeled sample data into the object detection model.

[0050] More preferably, in combination with a transfer learning strategy, the process of training the object detection model is accelerated, and the adaptability of the object detection model to data is improved. A multi-threaded parallel processing technology is adopted to detect multiple frames of images simultaneously.

[0051] More preferably, a federated learning strategy is adopted to train the object detection model to protect private data and improve the performance of the object detection model.

[0052] More preferably, after the object detection model is trained, the object detection model is evaluated by means of cross-validation and the like to ensure the stability and generalization ability of the model.

[0053] Preferably, the detection result generation module includes the following sub-modules:

[0054] The first sub-module for generating detection results is used to input the preprocessed image data into a lightweight neural network for further preliminary detection to form a preliminary detection result;

[0055] The second sub-module for generating detection results is used to input the preliminary detection result into the object detection model to form an object detection result.

[0056] More preferably, in the first sub-module for generating detection results, the lightweight neural network includes one of the YOLO-Tiny model, the MobileNet model, and the ShuffleNet model.

[0057] Preferably, in the judgment result generation module, a warning judgment is made according to a warning threshold and the object detection result.

[0058] More preferably, according to the warning threshold and the object detection result, a warning judgment is made in combination with a random forest (RF) classification model.

[0059] Preferably, a device for detecting edge - escaping wire of a circular shear further includes

[0060] a disk processing module, configured to determine whether early warning is required and process the wire jamming or edge - escaping wire of the circular shear according to the early warning judgment result. If required, give an early warning and process the wire jamming or edge - escaping wire of the circular shear;

[0061] a shearing parameter adjustment module, configured to adjust the circular shear parameters and / or the early warning threshold after processing the wire jamming or edge - escaping wire of the circular shear, and repeatedly execute the detection result generation module, the judgment result generation module, and the disk processing module until there is no abnormality in the circular shear.

[0062] In a third aspect, the present invention also provides a computer - readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the method for monitoring edge - escaping wire of a circular shear according to any one of the first aspects of the present application.

[0063] In a fourth aspect, the present invention also provides an electronic device. The electronic device includes: a memory storing a computer program; a processor communicatively connected to the memory, and when calling the computer program, executes the method for monitoring edge - escaping wire of a circular shear according to any one of the first aspects of the present application.

[0064] Compared with the prior art, the present invention has the following obvious outstanding substantial features and remarkable advantages:

[0065] The present invention provides a method and device for monitoring edge - escaping wire of a circular shear, which solves the problems of low accuracy and poor real - time performance in monitoring edge - escaping wire of a circular shear in the prior art. It has the following advantages: (1) Real - time and intelligent monitoring of edge - escaping wire and wire jamming of the circular shear is realized by collecting images in real time through a high - definition camera array and a 3D vision sensor; (2) The accuracy of judging edge - escaping wire and wire jamming is improved by combining deep - learning object detection algorithms, dynamic threshold setting, and deep - learning classification models; (3) The early warning threshold is dynamically set to ensure the timely transmission and processing of early warning information. The accuracy and real - time performance of monitoring edge - escaping wire of the circular shear are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] The accompanying drawings that form a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments and descriptions thereof of the present invention are used to explain the present invention and do not constitute an improper limitation to the present invention. In the drawings:

[0067] Figure 1 is a flowchart of a method for monitoring edge - escaping wire of a circular shear according to a preferred embodiment of the present invention.

[0068] Figure 2It is a schematic structural diagram of a monitoring device for escaping edge wires of a circular shear in a preferred embodiment of the present invention. Detailed implementation manners

[0069] The present invention provides a method and device for monitoring escaping edge wires of a circular shear. To make the objectives, technical solutions and effects of the present invention clearer and more definite, the following further describes the present invention in detail with reference to the accompanying drawings and by way of examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0070] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence. It should be understood that such data used in appropriate cases can be interchanged. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these process, method, product or device.

[0071] Example 1:

[0072] As Figure 1 shown, a method for monitoring escaping edge wires of a circular shear according to this embodiment specifically includes the following steps:

[0073] Step S1: Obtain image data for monitoring the escaping edge wire area of the circular shear.

[0074] Optionally, obtain image data of the area to be monitored through a high-definition camera array and a 3D vision sensor; wherein, the installation positions and installation angles of the high-definition camera array and the 3D vision sensor are optimized by a deep learning algorithm to ensure clear image capture under complex lighting and occlusion conditions.

[0075] Precisely install a high-definition camera array above the blanking area of the circular shear and combine it with a 3D vision sensor to ensure that their viewing angles can comprehensively and clearly cover the two blanking areas A and B. These cameras and sensors are configured to collect and transmit a fixed number of frames of images and depth information per second to a dedicated image processing computer. Intelligently optimize the viewing angles of the cameras to adapt to complex lighting and occlusion situations and ensure image quality.

[0076] Step S2: Preprocess the image data to form preprocessed image data.

[0077] Optionally, the preprocessing includes noise reduction; wherein, the image data is denoised by a Convolutional Neural Network (CNN). At the same time, through an adaptive contrast enhancement technique, the contrast is dynamically adjusted according to the image content to make the boundary between the steel material target and the background clearer.

[0078] Wherein, part of the data in the image data in step S1 is generated by GAN, or part of the preprocessed image data in step S2 is generated by GAN.

[0079] Step S3, construct an object detection model; wherein, the object detection model includes one or a combination of multiple of the R-CNN model, FPN model, YOLO series models, RetinaNet model, CenterNet model, EfficientDet model, DETR model. In this embodiment, the object detection model is the YOLO V7 model integrated with an attention mechanism.

[0080] Step S4, input the preprocessed image data into the object detection model to form an object detection result; wherein, the object detection result includes the shape of the steel material, the size of the steel material, the moving speed of the steel material, and the blanking result for monitoring the edge escape wire area of the circular shear.

[0081] Optionally, in step S4, before inputting the preprocessed image data into the object detection model, the object detection model is trained by inputting labeled sample data into the object detection model.

[0082] Wherein, in this embodiment, by combining a transfer learning strategy, the process of training the object detection model is accelerated, and the adaptability of the object detection model to data is improved. A multi-thread parallel processing technology is adopted to detect multiple frames of images simultaneously.

[0083] In this embodiment, the object detection model is trained by adopting a federated learning strategy to protect privacy data and improve the performance of the object detection model.

[0084] After the object detection model is trained, the object detection model is evaluated by means of cross-validation and the like to ensure the stability and generalization ability of the model.

[0085] Optionally, in step S4, before inputting the preprocessed image data into the object detection model, it further includes

[0086] Step S401: Input the preprocessed image data into a lightweight neural network for further preliminary detection to form a preliminary detection result. Among them, the lightweight neural network includes one of the YOLO-Tiny model, the MobileNet model, and the ShuffleNet model.

[0087] Step S402: Input the preliminary detection result into the target detection model to form a target detection result.

[0088] In addition, when performing target detection in step S4, the GPU parallel acceleration technology is adopted to further improve the speed of target detection.

[0089] Step S5: Make a warning judgment based on the target detection result to form a warning judgment result.

[0090] Among them, the warning judgment is made according to the warning threshold and the target detection result.

[0091] In this embodiment, the warning judgment is combined with the random forest (RF) classification model.

[0092] The system monitors the characteristics of the steel material such as shape, size, and speed in real time, and dynamically adjusts the warning threshold for the judgment of edge escape wire and block wire according to these parameters. For example, when the speed of the steel material increases, the warning threshold is correspondingly shortened to ensure the timeliness of the warning. When making a comprehensive judgment of multiple features, the deep learning algorithm long short-term memory network LSTM is introduced to predict the movement trajectory of the steel material, and a classification model for edge escape wire and block wire is constructed in combination with the random forest RF deep learning classification model. The model inputs include the characteristics of the steel material such as shape, size, and speed, as well as the detection results of two blanking areas A and B. The model is trained and evaluated through methods such as cross-validation to ensure the stability and generalization ability of the model. Within the set dynamic warning threshold, if no steel material is detected in blanking area A or B, and the classification model predicts a high risk of edge escape wire or block wire, the system automatically judges that there may be a phenomenon of block wire or edge escape wire in the disk shear and triggers the warning mechanism.

[0093] Step S6: According to the warning judgment result, judge whether warning is needed and deal with the block wire or edge escape wire of the disk shear. If needed, give a warning and deal with the block wire or edge escape wire of the disk shear.

[0094] Once the system determines that there is a phenomenon of edge wire escape or disc shear wire blockage, it immediately triggers the warning mechanism. This includes emitting an audible and visual alarm signal to attract the attention of the operator, and displaying the real-time image, depth information and related information of the edge wire escape or disc shear wire blockage phenomenon on the system interface. At the same time, the system will automatically record the relevant information, including the time point when the edge wire escape or wire blockage occurs, the specific location (area A or area B), and the characteristic data such as the shape, size and speed of the steel material. These information are crucial for the subsequent production process analysis and optimization, and the blockchain technology is used to ensure the immutability and security of the data.

[0095] In this embodiment, after receiving the warning information, measures should be immediately taken to deal with the phenomenon of edge wire escape or disc shear wire blockage.

[0096] Step S7: After dealing with the disc shear wire blockage or edge wire escape, the processing result is fed back through the user interface combined with natural language processing technology, and the disc shear parameters and / or warning thresholds are adjusted by the reinforcement learning method, and steps S4 - S6 are repeatedly executed until there is no abnormality in the disc shear.

[0097] Example 2:

[0098] As Figure 2 shown, a monitoring device for edge wire escape of a disc shear described in this embodiment specifically includes the following modules:

[0099] An image data acquisition module for acquiring image data for monitoring the edge wire escape area of the disc shear.

[0100] Among them, the image data of the area to be monitored is acquired through a high-definition camera array and a 3D vision sensor; the installation positions and installation angles of the high-definition camera array and the 3D vision sensor are optimized by a deep learning algorithm to ensure clear image capture under complex light and occlusion conditions.

[0101] An image data preprocessing module for preprocessing the image data to form preprocessed image data.

[0102] Among them, the preprocessing includes noise reduction, and the image data is processed for noise reduction through a convolutional neural network (CNN). In addition, through the adaptive contrast enhancement technology, the contrast is dynamically adjusted according to the image content to make the boundary between the steel material target and the background clearer.

[0103] Optionally, part of the data in the image data in the image data acquisition module is generated by GAN, or part of the data in the preprocessed image data in the image data preprocessing module is generated by GAN.

[0104] A model construction module for constructing a target detection model.

[0105] Among them, the object detection model includes one or a combination of multiple models such as the R-CNN model, the FPN model, the YOLO series models, the RetinaNet model, the CenterNet model, the EfficientDet model, and the DETR model. In this embodiment, the object detection model is the YOLO V7 model integrated with an attention mechanism.

[0106] The detection result generation module is configured to input the preprocessed image data into the object detection model to form an object detection result.

[0107] Among them, before inputting the preprocessed image data into the object detection model, the object detection model is trained by inputting labeled sample data into the object detection model.

[0108] In this embodiment, by combining a transfer learning strategy, the process of training the object detection model is accelerated, and the adaptability of the object detection model to data is improved. A multi-thread parallel processing technology is adopted to detect multiple frames of images simultaneously.

[0109] In this embodiment, the object detection model is trained by adopting a federated learning strategy to protect privacy data and improve the performance of the object detection model.

[0110] After the object detection model is trained, the object detection model is evaluated by means of cross-validation and the like to ensure the stability and generalization ability of the model.

[0111] Among them, the detection result generation module includes the following sub-modules:

[0112] The first sub-module for generating detection results is configured to input the preprocessed image data into a lightweight neural network for further preliminary detection to form a preliminary detection result; more preferably, in the first sub-module for generating detection results, the lightweight neural network includes one of the YOLO-Tiny model, the MobileNet model, and the ShuffleNet model.

[0113] The second sub-module for generating detection results is configured to input the preliminary detection result into the object detection model to form an object detection result.

[0114] The judgment result generation module is configured to make a warning judgment based on the object detection result to form a warning judgment result.

[0115] Among them, a warning judgment is made according to a warning threshold and the object detection result.

[0116] In this embodiment, according to the warning threshold and the target detection result, a warning judgment is made in combination with a random forest (RF) classification model.

[0117] The disk processing module is configured to judge whether a warning is needed and process the blocked wire or the escaping wire of the disk shear according to the warning judgment result. If a warning is needed, a warning is given and the blocked wire or the escaping wire of the disk shear is processed.

[0118] The shearing parameter adjustment module is configured to adjust the disk shearing parameters and / or the warning threshold after processing the blocked wire or the escaping wire of the disk shear, and repeatedly execute the detection result generation module, the judgment result generation module and the disk processing module until there is no abnormality in the disk shearing.

[0119] The specific embodiments of the present invention have been described in detail above, but they are only examples, and the present invention is not limited to the specific embodiments described above. For those skilled in the art, any equivalent modifications and substitutions made to the present invention are also within the scope of the present invention. Therefore, all equivalent transformations and modifications made without departing from the spirit and scope of the present invention should be covered within the scope of the present invention.

Claims

1. A method for monitoring the edge escape wire of a circular shear, characterized in that, Specifically, it includes the following steps: Step S1: Obtain image data for monitoring the edge escape wire area of the circular shear; Step S2: Preprocess the image data to form preprocessed image data; Step S3: Construct a target detection model; Step S4: Input the preprocessed image data into the target detection model to form a target detection result; wherein, the target detection result includes the shape of the steel material, the size of the steel material, the moving speed of the steel material, and the blanking result for monitoring the edge escape wire area of the circular shear; Step S5: Make a warning judgment based on the warning threshold and the target detection result to form a warning judgment result.

2. The edge trimming wire monitoring method for a circular shear according to claim 1, characterized in that In Step S2, the preprocessing includes: performing noise reduction processing on the image data through a convolutional neural network; and also dynamically adjusting the contrast according to the image content through an adaptive contrast enhancement technique to make the boundary between the steel material target and the background clearer.

3. The edge trimming wire monitoring method for a circular shear according to claim 1, characterized in that In Step S3, the target detection model includes one or a combination of multiple models among the R-CNN model, FPN model, YOLO series models, RetinaNet model, CenterNet model, EfficientDet model, and DETR model.

4. A method for monitoring the edge escape wire of a circular shear according to claim 1, characterized in that, In Step S4, before inputting the preprocessed image data into the target detection model, the target detection model is trained by inputting labeled sample data into the target detection model, including: Combining a transfer learning strategy to accelerate the training process of the target detection model and improve the adaptability of the target detection model to data, and adopting a multi-thread parallel processing technique to detect multiple frames of images simultaneously; Adopting a federated learning strategy to train the target detection model to protect privacy data and improve the performance of the target detection model; After the target detection model is trained, the target detection model is evaluated through a cross-validation method to ensure the stability and generalization ability of the model.

5. A method for monitoring the edge escape wire of a circular shear according to claim 1, characterized in that, In Step S4, before inputting the preprocessed image data into the target detection model, it also includes, Step S401: Input the preprocessed image data into a lightweight neural network for further preliminary detection to form a preliminary detection result; Step S402: Input the preliminary detection result into the target detection model to form a target detection result; Among them, in Step S401, the lightweight neural network includes one of the YOLO-Tiny model, MobileNet model, and ShuffleNet model.

6. A method for monitoring the edge escape wire of a circular shear according to claim 1, characterized in that, In Step S5, based on the warning threshold and the target detection result, a warning judgment is made in combination with a random forest classification model.

7. A method for monitoring the edge escape wire of a circular shear according to claim 1, characterized in that, It also includes, Step S6: According to the warning judgment result, judge whether it is necessary to give a warning and deal with the wire blockage or edge escape wire of the circular shear. If necessary, give a warning and deal with the wire blockage or edge escape wire of the circular shear; Step S7: After dealing with the wire blockage or edge escape wire of the circular shear, adjust the circular shear parameters and / or the warning threshold, and repeat Steps S4 - S6 until there is no abnormality in the circular shear.

8. A monitoring device for the edge escape wire of a circular shear, characterized in that, Specifically, it includes the following modules: An image data acquisition module for acquiring image data for monitoring the edge escape wire area of the circular shear; An image data preprocessing module for preprocessing the image data to form preprocessed image data; A model construction module for constructing a target detection model; A detection result generation module for inputting the preprocessed image data into the target detection model to form a target detection result; A judgment result generation module for making a warning judgment based on the target detection result to form a warning judgment result.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements a method for monitoring the edge escape wire of a circular shear according to any one of claims 1-7.

10. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a method for monitoring the edge escape wire of a circular shear according to any one of claims 1-7.

Images