Steel wire rope defect identification and positioning system and method based on real-time target detection

Abstract

The invention relates to the technical field of steel wire rope defect identification and positioning, in particular to a steel wire rope defect identification and positioning system and method based on real-time target detection. According to the technical scheme, the system comprises a sensor array, a synchronous detection execution mechanism, an NPU neural computing unit, a multi-modal weight management unit, a model library, a multi-core acquisition mainboard, a communication module and a power supply module. On the basis of a multi-modal real-time target detection method, the multi-modal cooperative training is realized by acquiring surface and internal key features of the steel wire rope through a multi-sensor array. The method not only improves the data characterization capability, optimizes the problems of model lightweight and boundary fuzziness, but also designs a self-defined dynamic allocation method, effectively improves the accuracy and precision of defect identification and positioning, realizes effective closed loop of model training, and provides a more efficient and reliable technical scheme for steel wire rope operation and maintenance enterprises.

Description

Technical Field

[0001] The present invention relates to the technical field of wire rope defect recognition and positioning, and in particular to a wire rope defect recognition and positioning system and method based on real-time target detection. Background Art

[0002] At present, the application of wire ropes in the fields of mines, ports, and construction is extremely important, involving the lifting, transportation, support, shock absorption, and safety assurance of equipment and personnel. Due to its high strength and wear resistance, wire ropes have become an indispensable key material in the field of equipment and personnel lifting. Considering that wire ropes are very prone to wear, rust, fatigue, wire breakage, and other problems during use, in severe cases, the wire ropes may break, causing losses to personnel and equipment. Under the existing model, wire rope defect detection mostly uses manual methods to detect ropes, which not only requires suspension of production and work, but also is labor-intensive and time-consuming, and the accuracy is not high. In order to solve the problems existing in the prior art, the present invention provides a wire rope defect identification and positioning method, system, and device based on multi-modal real-time target detection.

[0003] Therefore, the present application proposes a wire rope defect recognition and positioning system and method based on real-time target detection. Summary of the invention

[0004] The purpose of the present invention is to address the problem in the background technology that wire rope defect detection mostly adopts manual rope touching detection with low efficiency, and to propose a wire rope defect recognition and positioning system and method based on real-time target detection.

[0005] In a first aspect, the present application provides a wire rope defect recognition and positioning system based on real-time target detection, comprising: The sensor array includes a four-channel camera and a dual-channel magnetic probe, wherein the four-channel camera is used to collect planar images of the wire rope in four directions, and the dual-channel magnetic probe is used to obtain leakage magnetic field signals and magnetic flux signals; A synchronous detection actuator, used for driving the sensor array to move and clamp the steel wire rope for detection, and to release the steel wire rope and return to the initial position after the detection is completed; Neural Network Processing Unit (NPU) is used to accelerate the operation of deep learning models on collected data. A multimodal weight management unit is used to implement a multimodal dynamic allocation algorithm, to achieve dynamic allocation of weights and dynamic fusion of decisions; Model library, which is used to store the trained wire rope target detection model structure and parameter files, and supports online training and remote update; Multi-core acquisition motherboard, used to drive acquisition programs and detection programs, and implement operating system-level instruction operations and model training tasks; Communication module, supporting multiple communication methods; Power module, used to supply power to the entire system.

[0006] Optionally, the sensor array further includes: An image acquisition module, used to drive the four-channel camera to take photos and screenshots; Video acquisition module, used to continuously acquire video image information of the moving wire rope; An image processing unit, used to optimize the image resolution and display effect collected by the four-channel camera; An excitation mechanism, used to magnetize the wire rope; High-speed acquisition module, used to collect data collected by dual-channel magnetic probe; The signal conditioning circuit is used to denoise the data collected by the high-speed acquisition module.

[0007] Optionally, the NPU neural computing unit is externally mounted on a multi-core acquisition mainboard to perform real-time synchronous reasoning and analysis on camera images and magnetic probe data.

[0008] Optionally, the communication modes supported by the communication module include low-power wide area network communication (Lora) based on spread spectrum technology, low-power wide area network wireless communication technology (LPWAN), low-speed short-distance transmission technology Zigbee, narrowband Internet of Things technology (NB-IOT), 4G, and WIFI wireless communication modes.

[0009] In a second aspect, the present application provides a wire rope defect identification and positioning method based on real-time target detection, comprising the following steps: S1. Use four high-definition cameras to obtain videos and pictures of the wire rope in four directions: front, back, left and right, and perform image preprocessing and feature extraction; S2. Obtain the leakage magnetic field signal and magnetic flux signal of the wire rope through two magnetic probes, and perform signal conditioning and image conversion to obtain the leakage magnetic field signal modal diagram and magnetic flux modal diagram; S3, feature fusion of the camera image and the magnetic probe signal image at the same time to obtain a multi-channel image sample; S4, input the multi-channel image samples into the NPU neural computing unit for training to obtain detection model A, detection model B and detection model C; S5. Using a multimodal weight management unit, assigning modal weights to detection model A, detection model B, and detection model C; S6. Perform dynamic decision fusion on multi-channel image samples based on modal weight allocation and output wire rope defect location and identification results.

[0010] Optionally, S2 specifically includes: Use Hall and coil sensors to obtain the leakage magnetic field signal and magnetic flux signal of the wire rope; Condition the acquired signal and remove noise; The one-dimensional data is plotted and cut to obtain the magnetic field distribution, and then the leakage magnetic signal modal diagram and the magnetic flux modal diagram are obtained.

[0011] Optionally, in S4, the multi-channel image samples are input into the NPU neural computing unit, and are trained using a lightweight improved algorithm based on the yolov11 series target detection algorithm to obtain detection model A, detection model B and detection model C.

[0012] Optionally, S5 specifically includes: According to the multimodal dynamic allocation algorithm, based on the customized dynamic allocation method, modal weights are allocated to detection model A, detection model B, and detection model C; the contribution of the first ten principal components of the fusion modal graph of the wire rope camera is , the confidence level is , the corresponding multimodal model weight is ,but , the first ten principal components of the modal diagram of the wire rope magnetic flux leakage probe contribute to , the confidence level is , the first ten principal components of the modal diagram of the wire rope flux probe contribute to , the confidence level is , and the corresponding multimodal model weights are , ,but , , The final normalized weights are , , ,in, is the camera weight, is the camera normalization weight, is the magnetic flux leakage probe signal weight, is the normalized weight of the magnetic flux leakage probe signal, is the flux probe signal weight, Normalize weights for flux probe signals.

[0013] Optionally, the S6 specifically includes: According to the modal weight distribution results, dynamic decision fusion is performed on multi-channel image samples; Output wire rope defect location and identification results; The operating personnel conduct auxiliary verification of the two types of sensor data and call the synchronous detection actuator for back-check control to achieve secondary confirmation of the results.

[0014] Optionally, model optimization is also included, which specifically includes the following steps: The middle position of the wire rope defect is the original predicted position , by converting the prediction results into a sequence, it becomes: ,in, , To predict the label sequence, is the total number of sequences; by expanding and The probability of making the network quickly focus on the value near the label, and the loss of the wire rope target detection model is defined as: ,in, is the complete cross entropy value, is the predicted value, To predict the previous value, To predict the next value, we expand the probability of the values ​​on both sides of the boundary, that is, expand , ; , ,in, To predict the previous entropy value, To predict the next entropy value, ensure that the regression target is infinitely close to the true label of the corresponding wire rope defect position. ; According to the sample weight update formula, update the training sample weights so that in the next round of training, the model will pay more attention to the samples that were previously misclassified, thereby gradually improving the model performance. The sample weight update formula is: ,in, For the Wheel Sample The weight of For the Wheel Sample The weight of is the number of training rounds, Assign weights to the modes, is the true label, For sample The predicted label of , exp is the exponential function.

[0015] Compared with the prior art, the present invention has at least one of the following beneficial technical effects: The present invention is based on a multi-modal real-time target detection method, solves the problems existing in the field of wire rope defect identification and positioning, and achieves the following significant effects: The multi-sensor array fully acquires the key features of the surface and interior of the wire rope. The fusion of multi-modal data makes the data more comprehensive and accurate, providing a solid foundation for defect location and identification.

[0016] The multimodal collaborative training approach is used to optimize the lightweight and boundary fuzzy problems of the real-time detection model. This not only enables the model to process data more efficiently during training, but also reduces training time while ensuring model accuracy.

[0017] Aiming at the weight allocation problem of multimodal data training, a custom dynamic allocation method is designed. This method can reasonably allocate weights according to the characteristics and contribution of different modal data, thereby improving the accuracy and reliability of the model.

[0018] The multimodal joint analysis method overcomes the limitations of single modal data analysis and significantly improves the accuracy and precision of wire rope defect identification and positioning by comprehensively processing multiple modal data.

[0019] This system achieves effective feature learning of the wire rope defect recognition and positioning model by dynamically optimizing the model structure and training weights, forming a complete closed-loop system to ensure that the model can be continuously optimized and adapted to actual application needs.

[0020] In summary, the present invention is based on a multi-modal real-time target detection method, and collects key features of the surface and interior of the wire rope through a multi-sensor array to achieve multi-modal collaborative training. It not only improves the data characterization capability, optimizes the model lightweight and boundary fuzzy problems, but also designs a custom dynamic allocation method to effectively improve the accuracy and precision of defect identification and positioning, and realizes an effective closed loop of model training, providing a more efficient and reliable technical solution for wire rope operation and maintenance companies. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is the principle block diagram of the wire rope defect recognition and positioning system based on real-time target detection; Figure 2 The flowchart of the wire rope defect recognition and positioning method based on real-time target detection. DETAILED DESCRIPTION

[0022] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0023] See also Figure 1:The present application provides a wire rope defect identification and positioning system based on real-time target detection, which consists of a sensor array, a synchronous detection actuator, an NPU neural computing unit, a multi-modal weight management unit, a model library, a multi-core acquisition mainboard, a communication module, and a power module.

[0024] In this embodiment, the sensor array mainly includes a four-channel camera and a dual-channel magnetic probe. Among them, the four-channel sensor is used to collect planar images of the wire rope in four directions, and the defect information of the wire rope can be obtained through surface texture analysis. The image acquisition module is used to drive the camera to take pictures and screenshots, and the video acquisition module is used to continuously collect video image information of the moving wire rope. Considering the light environment and surface oil pollution problems of the wire rope, the image processing unit is used to optimize the image resolution and display effect. The dual-channel magnetic probe is divided into a Hall probe and a coil probe, which are used to obtain leakage magnetic field signals and magnetic flux signals respectively. Before the magnetic probe data is collected, the wire rope needs to be magnetized using an excitation mechanism, and then the high-speed acquisition module is used for data collection. The collected data is noisy and needs to be denoised using a signal conditioning circuit.

[0025] It is worth noting that the synchronous detection actuator is the front-end detection actuator of the equipment. After the mechanism is started, the motor drives the four-channel camera and the two-channel magnetic probe to move and clamp the wire rope, so that the wire rope runs through the inside of the equipment. After stopping the acquisition, the actuator releases the wire rope and moves back to the initial position to ensure the safe operation of the equipment.

[0026] Among them, the multi-core acquisition motherboard is used to drive the main device of the acquisition program and the detection program, and has a multi-core CPU and related general interfaces, such as the Raspberry Pi series core board or the Rockchip series core board, which can realize the instruction operation and model training tasks at the operating system level. By plugging in multiple NPU neural computing units, the accelerated operation of the deep learning model can be realized, and it can be used for real-time synchronous reasoning and analysis of camera images and magnetic probe data. In this embodiment, the multi-core acquisition motherboard has a multi-core central processing unit (CPU) and related general interfaces, such as the Raspberry Pi series core board or the Rockchip series core board, which can realize the instruction operation and model training tasks at the operating system level. It provides powerful computing support for the operation of the entire system, ensuring that data acquisition, processing and analysis can be carried out efficiently.

[0027] In addition, the multimodal weight management unit has a built-in multimodal dynamic allocation algorithm to achieve dynamic allocation of weights and dynamic fusion of decisions.

[0028] It should be noted that the model library, as a storage unit of the real-time target detection model, is used to store the trained wire rope target detection model structure and parameter files. At the same time, it supports functions such as online model training and remote update of the model library. The communication module supports a variety of communication mode options, such as Lora, Zigbee, NB-IOT, 4G, and WIFI wireless communication modes. It is convenient for data transmission and remote monitoring of the system. The power module supplies power to the entire device to ensure the stable operation of the system. This perfect hardware integration makes the system have good scalability and practicality. The power module is used to power the entire device.

[0029] See also Figure 2 , the present application provides a method for locating and identifying wire rope defects based on multi-modal real-time target detection, and the specific process is as follows. The wire rope passes through four high-definition cameras to obtain videos and pictures in four directions: front image, back image, left image and right image. After image preprocessing, high-quality images are obtained, and feature extraction is performed to mark the key parts of the image, etc., so as to obtain modal feature maps in four different directions. At the same time, the wire rope also passes through two magnetic probes, and the leakage magnetic field signal and magnetic flux signal are obtained by using Hall and coil sensors. After signal conditioning, a denoised signal is obtained, and the one-dimensional data is plotted and cut to obtain the magnetic field distribution, and the leakage magnetic signal modal map and the magnetic flux modal map are obtained. The two types of sensor arrays obtain images in four different directions and images drawn by two different magnetic signals, and the camera images at the same time are feature fused to obtain multi-channel image samples. After the NPU neural computing unit training model, the detection model A is obtained. At the same time, the image drawn by the magnetic signal at the same time, the unit performs feature extraction, and is trained using the NPU to obtain detection model B and detection model C respectively. Using the built-in allocation method of the multimodal weight management unit, based on the customized dynamic allocation method, the real-time reasoning weights of the three models A, B, and C are dynamically allocated, and the decisions are dynamically integrated to finally output the wire rope defect location and identification results. The operation and maintenance personnel can perform auxiliary verification based on the two types of sensor data, and finally determine the reliability of the wire rope defect type and location. In order to ensure the accuracy of the wire rope defect type and location, the synchronous detection actuator is called for back-check control to achieve secondary confirmation of the results.

[0030] As the core of multimodal real-time target detection of wire ropes, the model structure determines the accuracy of wire rope defect recognition and positioning. Here, based on the lightweight improved network (yolov11) series target detection algorithm, lightweight improvements are made, and GhostNetV3 is used to replace the original backbone network to reduce parameter complexity. At the same time, a neighborhood-aware self-attention mechanism is added to mine the key information of the feature map. Considering the boundary fuzziness problem of wire rope target detection, the defect boundary regression problem is converted into a classification problem, and the label prediction is converted into a sequence prediction. The middle position of the wire rope defect is the original predicted position. , by converting the prediction results into a sequence, it becomes ,in, , To predict the label sequence, is the total number of sequences.

[0031] By expanding and The probability of making the network quickly focus on the value near the label. Since the bounding box learning is only applicable to positive samples and there is no class imbalance problem, the loss of the wire rope target detection model is defined as ,in, is the complete cross entropy value, is the predicted value, To predict the previous value, To predict the next value. By expanding the probability of the values ​​on both sides of the boundary, that is, expanding , ; , ,in, To predict the previous entropy value, To predict the next entropy value, the regression target is infinitely close to the true label of the corresponding wire rope defect position. .

[0032] As the core of the dynamic allocation of multimodal weights for wire ropes, the customized dynamic allocation method determines the reliability of the wire rope defect identification and positioning results. The four modal images generated by the four-channel camera are placed in the detection model for inference after feature fusion to obtain the inference results. At the same time, the leakage magnetic modal image and the magnetic flux modal image generated by the dual-channel magnetic probe are each extracted and placed in the detection model for inference to obtain their respective inference results. Considering that the contribution of multimodal feature images to the results is different, the customized dynamic allocation method determines the weight coefficient based on the principal component feature ratio of the multimodal image and the iterative training of the multimodal image.

[0033] The first ten principal components of the fusion modal graph of the wire rope camera have a contribution of , the confidence level is , the corresponding multimodal model weight is ,but , the first ten principal components of the modal diagram of the wire rope magnetic flux leakage probe contribute to , the confidence level is , the first ten principal components of the modal diagram of the wire rope flux probe contribute to , the confidence level is ; The corresponding multimodal model weights are , ,but , , The final normalized weights are , , ,in, is the camera weight, is the camera normalization weight, is the magnetic flux leakage probe signal weight, is the normalized weight of the magnetic flux leakage probe signal, is the flux probe signal weight, Normalize the weights for the flux probe signal; Then, through iterative training, the training sample weights are updated, the weights of correctly classified samples are reduced, and the weights of incorrectly classified samples are increased. In this way, in the next round of training, the model will pay more attention to the previously incorrectly classified samples, thereby gradually improving the model performance. The sample weight update formula is as follows: ,in, For the Wheel Sample The weight of For the Wheel Sample The weight of is the number of training rounds, Assign weights to the modes, is the true label, For sample The predicted label of , exp is the exponential function.

[0034] At the same time, data set auxiliary verification has been added, which can intervene in the results through manual verification of labels and re-inspection of wire rope defect signals to achieve decision fusion and ultimately give the defect location and type.

[0035] It is worth noting that the image information and magnetic signal information of the wire rope are collected by a four-channel camera and a dual-channel magnetic probe respectively. This multi-modal data collection method can obtain the status information of the wire rope from different angles. Compared with the single-modal data collection, it can more comprehensively reflect the actual situation of the wire rope and reduce the misjudgment caused by the limitations of a single data type. The image features collected by the camera are fused with the magnetic signal features collected by the magnetic probe to obtain a multi-channel image sample. This fusion can comprehensively utilize the advantages of different modal data, dig out more features related to wire rope defects, and improve the accuracy of defect identification.

[0036] In this embodiment, considering the light environment and surface oil pollution of the wire rope, the image processing unit in the system can optimize the image resolution and display effect. This helps to obtain clear and usable image data in complex environments, ensuring that subsequent feature extraction and analysis can be carried out accurately.

[0037] In terms of model training, we made lightweight improvements based on the Yolov11 series target detection algorithm, and used GhostNetV3 to replace the original backbone network, reducing the complexity of parameters. This not only improves the operational efficiency of the model, but also speeds up the detection speed without losing too much accuracy, meeting the needs of real-time detection. Adding a neighborhood-aware self-attention mechanism can mine key information from feature maps, further improving the model's ability to extract wire rope defect features, and helping to identify defects more accurately.

[0038] Furthermore, the multimodal weight management unit can dynamically allocate weights according to the contribution of different modal data to the results through the built-in multimodal dynamic allocation algorithm. This dynamic allocation mechanism can adaptively adjust the importance of each modal data in decision-making and improve the adaptability of the entire system to different environments and wire rope states. Based on weight allocation, dynamic decision fusion of multi-channel image samples can comprehensively consider the analysis results of each modality, avoid the one-sidedness of single modal decision-making, and thus output wire rope defect location and identification results more accurately.

[0039] In this embodiment, the operation and maintenance personnel can perform auxiliary verification based on the two types of sensor data, and perform back-check control by calling the synchronous detection actuator to achieve secondary confirmation of the results. This multiple verification mechanism can effectively improve the reliability of defect identification results and ensure the accuracy of the detection results. Through the sample weight update formula, during the iterative training process, the weight of correctly classified samples is reduced, while the weight of incorrectly classified samples is increased. This allows the model to pay more attention to previously misclassified samples in subsequent training, gradually improve model performance, and continuously improve the accuracy of defect identification.

[0040] The above specific embodiments are only several optional embodiments of the present invention. Based on the technical solutions of the present invention and the relevant inspirations of the above embodiments, those skilled in the art can make various alternative improvements and combinations to the above specific embodiments.

Claims

1. A wire rope defect recognition and positioning system based on real-time target detection, characterized in that: include: The sensor array includes a four-channel camera and a dual-channel magnetic probe, wherein the four-channel camera is used to collect planar images of the wire rope in four directions, and the dual-channel magnetic probe is used to obtain leakage magnetic field signals and magnetic flux signals; A synchronous detection actuator, used for driving the sensor array to move and clamp the steel wire rope for detection, and to release the steel wire rope and return to the initial position after the detection is completed; NPU neural computing unit, used to accelerate the operation of deep learning models on collected data; A multimodal weight management unit is used to implement a multimodal dynamic allocation algorithm, to achieve dynamic allocation of weights and dynamic fusion of decisions; Model library, which is used to store the trained wire rope target detection model structure and parameter files, and supports online training and remote update; Multi-core acquisition motherboard, used to drive acquisition programs and detection programs, and implement operating system-level instruction operations and model training tasks; Communication module, supporting multiple communication methods; Power module, used to supply power to the entire system.

2. The wire rope defect identification and positioning system based on real-time target detection according to claim 1 is characterized in that: The sensor array further comprises: An image acquisition module, used to drive the four-channel camera to take photos and screenshots; Video acquisition module, used to continuously acquire video image information of the moving wire rope; An image processing unit, used to optimize the image resolution and display effect collected by the four-channel camera; An excitation mechanism, used to magnetize the wire rope; High-speed acquisition module, used to collect data collected by dual-channel magnetic probe; The signal conditioning circuit is used to denoise the data collected by the high-speed acquisition module.

3. The wire rope defect identification and positioning system based on real-time target detection according to claim 1 is characterized in that: The NPU neural computing unit is externally mounted on a multi-core acquisition motherboard and is used for real-time synchronous reasoning and analysis of camera images and magnetic probe data.

4. The wire rope defect identification and positioning system based on real-time target detection according to claim 1, characterized in that: The communication modes supported by the communication module include Lora, Zigbee, NB-IOT, 4G, and WIFI wireless communication modes.

5. A wire rope defect recognition and positioning method based on real-time target detection, characterized in that: The following steps are involved: S1. Use four high-definition cameras to obtain videos and pictures of the wire rope in four directions: front, back, left and right, and perform image preprocessing and feature extraction; S2. Obtain the leakage magnetic field signal and magnetic flux signal of the wire rope through two magnetic probes, and perform signal conditioning and image conversion to obtain the leakage magnetic field signal modal diagram and magnetic flux modal diagram; S3, feature fusion of the camera image and the magnetic probe signal image at the same time to obtain a multi-channel image sample; S4, input the multi-channel image samples into the NPU neural computing unit for training to obtain detection model A, detection model B and detection model C; S5. Using a multimodal weight management unit, assigning modal weights to detection model A, detection model B, and detection model C; S6. Perform dynamic decision fusion on multi-channel image samples based on modal weight allocation and output wire rope defect location and identification results.

6. The method for identifying and locating wire rope defects based on real-time target detection according to claim 5 is characterized in that: The S2 specifically includes: Use Hall and coil sensors to obtain the leakage magnetic field signal and magnetic flux signal of the wire rope; Condition the acquired signal and remove noise; The one-dimensional data is plotted and cut to obtain the magnetic field distribution, and then the leakage magnetic signal modal diagram and the magnetic flux modal diagram are obtained.

7. The method for identifying and locating wire rope defects based on real-time target detection according to claim 5 is characterized in that: In S4, the multi-channel image samples are input into the NPU neural computing unit, and are trained using a lightweight improved algorithm based on the yolov11 series target detection algorithm to obtain detection models A, B and C.

8. The method for identifying and locating wire rope defects based on real-time target detection according to claim 5, characterized in that: The S5 specifically includes: According to the multimodal dynamic allocation algorithm, based on a custom dynamic allocation method, modal weights are allocated to detection model A, detection model B, and detection model C; The first ten principal components of the fusion modal graph of the wire rope camera have a contribution of The confidence level is The corresponding multimodal model weights are ;but , the first ten principal components of the modal diagram of the wire rope magnetic flux leakage probe contribute to , the confidence level is ; The first ten principal component contributions of the modal diagram of the wire rope flux probe are , the confidence level is , and the corresponding multimodal model weights are , ,but ， ， The final normalized weights are ， ， ， in, is the camera weight, is the camera normalization weight, Magnetic flux leakage probe signal weight, is the normalized weight of the magnetic flux leakage probe signal, is the flux probe signal weight, Normalize weights for flux probe signals.

9. The method for identifying and locating wire rope defects based on real-time target detection according to claim 8, characterized in that: The S6 specifically includes: According to the modal weight distribution results, dynamic decision fusion is performed on multi-channel image samples; Output wire rope defect location and identification results; The operating personnel conduct auxiliary verification of the two types of sensor data and call the synchronous detection actuator for back-check control to achieve secondary confirmation of the results.

10. The method for identifying and locating wire rope defects based on real-time target detection according to claim 9, characterized in that: It also includes model optimization, which includes the following steps: The middle position of the wire rope defect is the original predicted position , by converting the prediction results into a sequence, it becomes: ， in, , To predict the label sequence, is the total number of sequences; By expanding and The probability of making the network quickly focus on the value near the label, and the loss of the wire rope target detection model is defined as: ， in, is the complete cross entropy value, is the predicted value, To predict the previous value, To predict the next value, we expand the probability of the values ​​on both sides of the boundary, that is, expand , ; ， ， in, To predict the previous entropy value, To predict the next entropy value, ensure that the regression target is infinitely close to the true label of the corresponding wire rope defect position ; According to the sample weight update formula, the training sample weights are updated so that in the next round of training, the model will pay more attention to the samples that were previously misclassified, thereby gradually improving the model performance. The sample weight update formula is: ， in, For the Wheel Sample The weight of For the Wheel Sample The weight of is the number of training rounds, Assign weights to the modes, is the true label, For sample The predicted label of , exp is the exponential function.

Images