Crown block safety monitoring system based on YOLO V5 and Wincc

Abstract

The invention discloses a crown block safety monitoring system based on YOLO V5 and Wincc, belongs to the field of monitoring systems, and aims to solve the problems that a crown block monitoring system mainly depends on manual judgment and is easily influenced by experience and judgment of an operator, and the operation risk is increased. The system comprises a data acquisition module, an image vision module, a hoisting state judgment module and an abnormity alarm module, the data acquisition module is used for acquiring variable point locations of the crown block PLC in real time; the method comprises the following steps: reading a variable point location in a PLC (Programmable Logic Controller) through KepServerEX software; the image vision module is used for recording a steel ladle trunnion position working image in real time from the front angle through n high-definition network cameras mounted on the two sides of a crown block working area and transmitting the steel ladle trunnion position working image to the hoisting state judgment module; the hoisting state judgment module is used for receiving the steel ladle trunnion position working image transmitted from the image vision module; and performing visual image identification on the steel ladle trunnion position working image based on a YOLO V5 algorithm, and identifying whether a lifting lug of the steel ladle and a lifting hook of the crown block are in safe positions or not.

Description

Technical Field

[0001] The present invention relates to a monitoring system, and particularly to a crane safety monitoring system based on YOLO V5 and Wincc. Background Art

[0002] In traditional crane monitoring systems, manual judgment is mainly relied on. Especially in the inspection of the hooking state between the hook and the ladle trunnion, it is easily affected by the operator's experience and judgment ability, resulting in misjudgment or slow response, thus increasing the operation risk. Although vision algorithms can assist in manually judging the hooking state between the hook and the trunnion, environmental factors such as light and soot may affect the accuracy of visual detection, resulting in misjudgment or delay. In addition, the performance of the system depends on a stable network environment, and data transmission and network problems may affect the real-time data display of WinCC and the response speed of the vision algorithm.

[0003] In addition, the height and weight data collected by sensors may be affected by environmental factors such as vibration and temperature, resulting in inaccurate or unstable data. The network transmission frequency of the existing system is low, and the real-time performance of data update and display is poor, unable to effectively meet the production requirements of high speed and high load, resulting in lag or absence of monitoring information, affecting operation safety and efficiency. Especially during data transmission, the existing network cannot provide sufficient bandwidth and low latency, restricting the ability of real-time monitoring and response, and further increasing the production risk. Although the system has integrated a variety of advanced technologies, its robustness and stability still need to be further optimized in complex environments. Summary of the Invention

[0004] Aiming at the deficiencies of the existing technology, the present invention proposes a crane safety monitoring system based on YOLO V5 and Wincc, including a data acquisition module, an image vision module, a lifting state judgment module, and an abnormal alarm module; The data acquisition module collects the variable points of the crane PLC in real time; reads the variable points in the PLC through KepServerEX software; The image vision module records the working image of the ladle trunnion position in real time from the front angle through n high-definition network cameras installed on both sides of the crane working area and transmits it to the lifting state judgment module; The lifting state judgment module receives the working image of the ladle trunnion position transmitted from the image vision module; based on the YOLO V5 algorithm, it performs visual image recognition on the working image of the ladle trunnion position to identify whether the positions of the ladle's lifting lugs and the crane's hook are in a safe position: performs machine vision recognition on the ladle trunnion and the hook, calculates the distance between the midpoints of the diagonals of the hook rectangular frame and the ladle trunnion rectangular frame, and judges whether the current ladle trunnion and the hook are firmly hooked based on the distance value between the two points. If the distance value is greater than a specific value, the judgment result is not firmly hooked, and an abnormal signal of not being firmly hooked is sent to the abnormal alarm module; The abnormal alarm module performs OPC communication between KepServerEX and Wincc; Wincc reads the variable point data in KepServerEX, and then reads the real-time variable point data in the PLC; sets an alarm indicator light in the configuration screen of Wincc. When an abnormal point in the PLC is triggered, the alarm indicator light turns red; receives the abnormal signal transmitted from the lifting state judgment module, sends the abnormal signal to flash on the alarm light, and the lifting state indicator light set in the Wincc screen turns red and flashes.

[0005] As a further improvement of the present invention, during the data acquisition process, the variable points in the PLC are read through the KepServerEX software, including the weight of the hooked object, the position of the crane, the height of the hook, the operating gear of the crane, and the alarm points, namely overload, overcurrent, and limit.

[0006] As a further improvement of the present invention, there are two high-definition network cameras.

[0007] As a further improvement of the present invention, the method for specifically identifying whether the positions of the ladle's lifting lugs and the crane's hook are in a safe position in the lifting judgment module includes the following methods: First, the input layer receives the working image of the ladle trunnion position collected by the camera in real time. When converting the 1280×960 pixel image to 608×608, the calculated width scaling factor is 0.475, and the height scaling factor is 0.633. Then, the smaller scaling factor of 0.475 is selected for scaling to change the original image size to 608×456. Next, 76 black areas are filled above and below the image to meet the input requirement of 608×608. Subsequently, through the convolutional layer and stride, downsampling is achieved to gradually reduce the spatial resolution of the feature map while increasing the channels of the feature map. Multiple convolutional layers are used to extract features and output the feature map. Batch Normalizaton is added after the convolutional layer to accelerate training and improve the stability of the model. Next, the output feature map is divided into two parts by CSP. One part undergoes further convolutional operations, and the other part is directly passed to the next layer. Finally, the processed features and the unprocessed features are merged to form a new feature map. The ladle, ladle trunnion, and hook in the new feature map are further identified. When the hook rectangular box is recognized, the distance between the midpoint of the diagonal of the hook rectangular box and the midpoint of the diagonal of the ladle trunnion rectangular box is taken. If the distance between the two points is greater than a specific value, it is determined that the current state is not hooked firmly.

[0008] As a further improvement of the present invention, data exchange with the overhead crane PLC is realized between the data acquisition module, the image vision module, the lifting state judgment module, and the abnormal alarm module through the 5G CPE and the OPC protocol.

[0009] As a further improvement of the present invention, in the lifting state judgment module, in order to further confirm the judgment result, the ladle weight is introduced as an auxiliary judgment means.

[0010] The beneficial effects of the present invention are as follows: 1. By introducing the YOLO V5 vision algorithm to assist manual judgment, the problems of manual error and reaction delay in the judgment of the hooking state between the hook and the ladle trunnion are solved, and the accuracy and real-time performance of the judgment are significantly improved. At the same time, with the help of the 5G CPE device, the system realizes low-latency and high-bandwidth network transmission, greatly improving the data update frequency and response speed, and ensuring real-time monitoring and data transmission in a high-load production environment.

[0011] 2. Using the OPC protocol, data collection and exchange with the overhead crane PLC are carried out through the KepServerEX software, ensuring that the key parameters of the overhead crane (such as position, height, load, etc.) can be transmitted and displayed in real time and accurately. The overall performance and stability are greatly improved, thus effectively ensuring operation safety and production efficiency.

[0012] 3. Integrate industrial automation control and computer vision technology to conduct real-time monitoring and intelligent judgment on the operating status of the overhead crane. Adopt the OPC protocol to collect and exchange data with the overhead crane PLC through the KepServerEX software. KepServerEX is used for data collection and transmission. The key parameters of the overhead crane are obtained through sensors and displayed in real time on the Wincc platform. The YOLO V5 algorithm is applied to the visual recognition of the ladle trunnion and the overhead crane hook, automatically detecting whether the hook is correctly hooked and tracking its operating trajectory, replacing the traditional manual judgment method.

[0013] 4. Implement centralized monitoring of the overhead crane through the Wincc platform. The operator can view the device status and alarm information in real time, thus ensuring operation safety. The system provides a network connection with low latency and high bandwidth through 5G CPE devices, significantly improving the data update frequency and response speed. Using the OPC protocol, the system exchanges data with the overhead crane PLC to obtain key parameters such as position information and load in real time, ensuring the accuracy and timeliness of the monitoring data. KepServerEX, as an OPC server, optimizes data collection and transmission. The visual algorithm assists manual judgment of the hooking status of the hook and the ladle trunnion, improving the monitoring accuracy and efficiency, and providing more accurate decision-making support for the operator. Brief Description of the Drawings

[0014] Figure 1 is the structural diagram of the present invention; Figure 2 is the YOLO V5 network structural diagram of the present invention; Figure 3 is the monitoring diagram of the ladle working station of the present invention; Detailed Embodiments

[0015] To facilitate the understanding of this application, the following will provide a more comprehensive description of this application with reference to the relevant drawings. The embodiments described are shown in the drawings. On the contrary, the purpose of providing these embodiments is to make the disclosure of this application more thoroughly understood.

[0016] In this embodiment, the camera selected is the DH-IPC-HFW1230M-A-I1-V5 Dahua 2 million infrared high-definition network bullet camera, with a camera bit rate of 1920×1080@25fps and equipped with a dust and shock-proof housing. The housing has an air-cooling interface. If the on-site temperature is relatively high, the user can connect the on-site compressed air; In this embodiment, the processor is an Intel® Core™ i7-10750H CPU@2.6GHz processor with 16GB of memory.

[0017] The overhead crane safety monitoring system based on YOLO V5 and Wincc combines industrial automation control and computer vision technology to achieve real-time monitoring and intelligent judgment of the operating status of the overhead crane. The system uses a 5G CPE device, the OPC protocol, KepServerEX software, and the YOLO V5 vision algorithm to achieve real-time collection and transmission of overhead crane PLC data, as well as visual recognition of the ladle trunnion and the overhead crane hook, thereby improving operation safety, reliability, and efficiency.

[0018] As Figure 1 shown, an overhead crane safety monitoring system based on YOLO V5 and Wincc includes a data acquisition module, an image vision module, a lifting state judgment module, and an abnormal alarm module; The data acquisition module collects the variable points of the overhead crane PLC in real time and transmits them to the abnormal alarm module. The variable points include the weight of the hooked object, the position of the overhead crane, the height of the hook, the gear of the overhead crane operation, and alarm points (overload, overcurrent, limit); During the data acquisition process, the variable points in the PLC are read through KepServerEX software, including the weight of the hooked object, the position of the overhead crane, the height of the hook, the gear of the overhead crane operation, and alarm points such as overload, overcurrent, and limit.

[0019] As Figure 2 shown, the image vision module uses 2 high-definition network cameras installed on both sides of the overhead crane working area to record the working image of the ladle trunnion position from the front angle in real time and transmits it to the lifting state judgment module; The lifting state judgment module receives the working image of the ladle trunnion position transmitted from the image vision module; based on the YOLO V5 algorithm, it performs visual image recognition on the working image of the ladle trunnion position to identify whether the position of the ladle ear and the overhead crane hook is in a safe position: performs machine vision recognition on the ladle trunnion and the hook, calculates the distance between the midpoints of the diagonals of the hook rectangle frame and the ladle trunnion rectangle frame, and judges whether the current ladle trunnion and the hook are firmly hooked through the distance value between the two points, and introduces the ladle weight as an auxiliary judgment means; if the distance value is greater than a specific value, the judgment result is not firmly hooked, and the abnormal signal of not being firmly hooked is sent to the abnormal alarm module; Specifically, the judgment of the overhead crane lifting state is achieved through the following process: First, the working image of the ladle trunnion position is connected to the network video encoder through HDMI, and then through the decoder installed on the overhead crane, the video stream is received and decoded using the 5G CPE network of the overhead crane, and finally it is displayed on the monitor observed by the overhead crane operator through HDMI to provide auxiliary judgment information for the overhead crane driver; Meanwhile, according to the on-site environment, a network camera adapted to the resolution of the current working condition is installed in the hard disk video recorder. The inference server calls the video stream through the hard disk video recorder and uses the YOLO V5 open-source framework model to perform visual image recognition on the video stream to identify the positions of the ladle lugs and the crane hook. On this basis, through the statistical analysis of a large number of image recognition features, the appropriate algorithm range is determined, and then the positions of the ladle lugs and the crane hook are judged and a prompt is given; Finally, the inference server connects the rendered image to the network video encoder through HDMI. The decoder installed on the crane receives the video stream through the 5G CPE network of the crane, decodes it, and then displays it on the monitor observed by the crane operator through HDMI. Through this series of operations, the system can accurately identify the positions of the ladle lugs and the crane hook, and provide instant visual information for the operator, thereby improving work efficiency and ensuring operation safety.

[0020] In June 2020, Jocher proposed YOLO V5. Compared with previous versions, YOLO V5 has been improved and optimized in many aspects. Compared with previous versions of YOLO, the most significant improvement is the fast detection speed of YOLO V5. The inference time for a single image can reach 0.007s. Thanks to the small-size model of YOLO V5, it can be quickly deployed in various usage scenarios. Taking YOLO V5 as an example, its network structure is shown in the figure. This structure is divided into an input end, a backbone network Backbone, a Neck network, and an output end Prediction.

[0021] The YOLO V5 model usually accepts input images with sizes of 416×416 or 608×608. For images of different sizes, they will be scaled proportionally, and when the height and width are different, black areas will be automatically filled on both sides to match the input size.

[0022] In the YOLO V5 model, the design highlights of the backbone network lie in the Focus and CSP structures. Through slicing and convolution operations, they efficiently transform the image into a compact feature map, thereby significantly reducing the complexity and computational cost of the model while maintaining high accuracy, and achieving the lightweight of the model.

[0023] As Figure 3 shown, compared with other YOLO versions, the Neck network of YOLO V5 adopts a lightweight feature fusion module PAN (Path Aggregation Network). PAN consists of an upsampling module and a fusion module. These two components can effectively fuse feature maps from different levels, thereby obtaining richer context information, improving detection performance, and enhancing robustness.

[0024] When constructing a target detection algorithm, the design of the loss function is crucial, mainly including two major elements: classification loss and regression loss. When evaluating the overlapping degree between the predicted bounding box and the ground truth bounding box, the Intersection over Union (IOU) is commonly used as an indicator, which reflects the ratio of the intersection area to the union area of the predicted bounding box and the ground truth bounding box. However, the Loss of IOU has two problems: one is that when the predicted bounding box and the ground truth bounding box do not overlap at all, the IOU value is 0, resulting in the loss function being unable to be effectively differentiated and optimized; the other is that even if the areas of the predicted bounding box and the ground truth bounding box are the same, the intersection situations may be very different, but IOU_Loss cannot distinguish these situations.

[0025] To solve these problems, YOLO V5 adopts a more refined CIOU_Loss. CIOU_Loss not only considers the overlapping region between the predicted bounding box and the ground truth bounding box, but also takes into account two key factors: the aspect ratio and the center point position, thus achieving a more comprehensive and accurate evaluation.

[0026] Specifically, the image annotation software LabelImg is used to annotate the main hook of the crane and the trunnion of the ladle. Labelimg is a labeling tool that can label multiple categories and directly generate txt files. Its function is to label the position of the target object in the original image and generate a corresponding txt file for each image to represent the position of the target standard bounding box. In the project, the YOLO label format is used, and the labeled labels are stored in the txt file. After annotation, model training is carried out. The recognition algorithm for the main hook of the crane and the trunnion of the ladle based on the YOLO V5 algorithm runs on the processor, and the deep learning framework is Pytorch.

[0027] To improve the recognition accuracy, a large number of image data are captured from the hard disk video recorder for training. The inference server uses the hard disk video recorder to call the video stream and applies the YOLO V5 open-source framework model for visual image recognition. The trained model can accurately identify the lifting lugs of the ladle and the hook of the crane. By statistically analyzing a large number of image recognition features, the algorithm range suitable for the current working conditions can be determined and accurate judgments can be made.

[0028] Considering the possible influence of on-site light conditions, pictures were captured at different times, with a total of 1600 pictures and a resolution of 1280×960. They were randomly divided in an 8:2 ratio, with 1280 pictures as the training set and 320 as the test set. The same is true for YOLO V5. The Backbone part is the feature extraction network of the model, responsible for extracting useful feature information from the input image, that is, the working image of the ladle trunnion position. YOLO V5 uses a structure that combines the Feature Pyramid Network (FPN) and the Path Aggregation Network (PAN). The Head part converts the feature map output by the Neck into the final bounding boxes and class predictions. Overall, the Backbone is responsible for feature extraction, the Neck is used to aggregate features from different layers of the Backbone for object detection at multiple scales, and the Head is responsible for generating the final detection results. Such a hierarchical design enables YOLO V5 to achieve good detection performance while maintaining high efficiency, effectively improving the model's expressive ability and computational efficiency; The specific methods for feature extraction of the working image of the ladle trunnion position are as follows: First, the input layer receives the working image of the ladle trunnion position captured in real time by the camera. When converting a 1280×960 pixel image to 608×608, first calculate the width scaling factor of 0.475 and the height scaling factor of 0.633, then select the smaller scaling factor of 0.475 for scaling to change the original image size to 608×456, and then fill 76 black areas at the top and bottom of the image to meet the input requirement of 608×608; Subsequently, through convolutional layers and strides, downsampling is implemented to gradually reduce the spatial resolution of the feature map while increasing the number of channels of the feature map. Multiple convolutional layers are used to extract features and maintain spatial information, and the feature map is output; The LeakReLU function is used to increase non-linear features and learn more complex features; Batch Normalization is added after the convolutional layer to accelerate training and improve the stability of the model; Next, the output feature map is split into two parts by CSP. One part undergoes further convolutional operations, including additional BN and activation, and the other part is directly passed to the next layer without additional processing, similar to the skip connection in the ResNet network. This design is actually gradually improved based on ResNet: first, DenseNet is improved from ResNet, and then further improved to CSP (Cross-Stage Partial connections). Finally, the processed features and the unprocessed features are merged to form a new feature map; The new feature map formed by the above method further identifies the ladle, ladle trunnion, and hook, and identifies the distance between the midpoint of the diagonal of the hook rectangular frame and the midpoint of the diagonal of the ladle trunnion rectangular frame. If the distance between the two points is greater than a specific value, it is determined that the current is not hooked firmly. To avoid misjudgment, the ladle weight is introduced for auxiliary judgment.

[0029] The abnormal alarm module conducts OPC communication between KepServerEX and Wincc; Wincc reads the variable point data in KepServerEX, and then reads the real-time variable point data in the PLC; an alarm indicator light is set on the configuration screen of Wincc. When an abnormal point in the PLC is triggered, the alarm indicator light turns red to remind the on-site staff; it receives the abnormal signal transmitted by the lifting state judgment module and sends the abnormal signal to flash on the alarm light, and the lifting state indicator light set on the Wincc screen turns red and flashes.

[0030] The overhead crane walks into work station A within the hanging area, and transmits the real-time video signal to the intelligent display device in the cab through the wireless base station and immediately displays the video of work station A; when it enters B, it transmits the real-time video signal to the intelligent display device in the cab through the wireless base station and immediately displays the video of work station B. The PLC data of nine overhead cranes is collected through KepServerEX (a total of more than 1,000 pieces of data), and then OPC communication is carried out between KepServerEX and Wincc. Wincc reads the data collected in KepServerEX in real time, and the operating parameters of the overhead crane are displayed on Wincc. Then, support for WEB is set on Wincc, and finally, real-time monitoring of the overhead crane PLC data can be realized on the WEB side to provide early warnings for on-site workers.

[0031] By introducing the YOLO V5 vision algorithm to assist manual judgment, the problems of manual error and reaction delay in judging the hooking state between the hook and the ladle trunnion are solved, and the accuracy and real-time performance of the judgment are significantly improved. At the same time, with the help of 5G CPE equipment, the system realizes low-latency and high-bandwidth network transmission, greatly improving the data update frequency and response speed, and ensuring real-time monitoring and data transmission in a high-load production environment.

[0032] Adopting the OPC protocol, the data of the overhead crane PLC is collected and exchanged through KepServerEX software, ensuring that the key parameters of the overhead crane (such as position, height, load, etc.) can be transmitted and displayed in real time and accurately. The overall performance and stability are greatly improved, thus effectively ensuring operation safety and production efficiency.

[0033] Combined with industrial automation control and computer vision technology, the operating status of the overhead crane is monitored in real time and judged intelligently. Data exchange with the overhead crane PLC is achieved through a 5G CPE device using the OPC protocol, and KepServerEX is used for data acquisition and transmission. The key parameters of the overhead crane are obtained through sensors and displayed in real time on the Wincc platform. The YOLO V5 algorithm is applied to the visual recognition of the ladle trunnion and the overhead crane hook, automatically detecting whether the hook is correctly hooked and tracking its running trajectory, replacing the traditional manual judgment method.

[0034] Centralized monitoring of the overhead crane is realized through the Wincc platform, and operators can view the device status and alarm information in real time, thus ensuring operation safety. The system provides a network connection with low latency and high bandwidth through a 5G CPE device, significantly improving the data update frequency and response speed. Using the OPC protocol, the system exchanges data with the overhead crane PLC to obtain key parameters such as position information and load in real time, ensuring the accuracy and timeliness of the monitoring data. As an OPC server, KepServerEX optimizes data acquisition and transmission. The vision algorithm assists manual judgment of the hooking status of the hook and the ladle trunnion, improving the monitoring accuracy and efficiency, and providing more accurate decision-making support for operators.

[0035] The above-described embodiments merely represent several implementation manners of the present invention, and their descriptions are relatively specific and detailed, but should not be construed as limiting the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention patent shall be subject to the appended claims.

Claims

1. A crane safety monitoring system based on YOLO V5 and Wincc, characterized in that It includes a data acquisition module, an image vision module, a lifting state judgment module, and an abnormal alarm module; The data acquisition module collects the variable points of the overhead crane PLC in real time; reads the variable points in the PLC through the KepServerEX software; The image vision module records the working image of the ladle trunnion position in real time from the front angle through n high-definition network cameras installed on both sides of the working area of the overhead crane, and transmits it to the lifting state judgment module; The lifting state judgment module receives the working image of the ladle trunnion position transmitted from the image vision module; performs visual image recognition on the working image of the ladle trunnion position based on the YOLOV5 algorithm to identify whether the position of the ladle's lifting lug and the overhead crane's hook is in a safe position: performs machine vision recognition on the ladle trunnion and the hook, calculates the distance between the midpoints of the diagonals of the hook rectangle frame and the ladle trunnion rectangle frame, and judges whether the current ladle trunnion and the hook are firmly hooked through the distance value between the two points. If the distance value is greater than a specific value, the judgment result is not firmly hooked, and the abnormal signal of not being firmly hooked is sent to the abnormal alarm module; The abnormal alarm module conducts OPC communication between KepServerEX and Wincc; Wincc reads the variable point data in KepServerEX, and then reads the real-time data of the variable points in the PLC; sets an alarm indicator light in the configuration screen of Wincc. When an abnormal point in the PLC is triggered, the alarm indicator light turns red; receives the abnormal signal transmitted from the lifting state judgment module, sends the abnormal signal to flash on the alarm light, and the lifting state indicator light set in the Wincc screen turns red and flashes.

2. The overhead crane safety monitoring system based on YOLO V5 and Wincc according to claim 1, characterized in that During the data acquisition process, the variable points in the PLC are read through the KepServerEX software, including the weight of the hooked object, the position of the overhead crane, the height of the hook, the operating gear of the overhead crane, and the alarm points, namely overload, overcurrent, and limit.

3. The overhead crane safety monitoring system based on YOLO V5 and Wincc according to claim 1, characterized in that, There are two high-definition network cameras.

4. The overhead crane safety monitoring system based on YOLO V5 and Wincc according to claim 1, characterized in that, The specific method for the lifting judgment module to identify whether the position of the ladle's lifting lug and the overhead crane's hook is in a safe position includes the following: First, the input layer receives the working image of the ladle trunnion position collected in real time by the camera. When converting a 1280×960 pixel picture to 608×608, the calculated width scaling factor is 0.475, and the height scaling factor is 0.

633. Then, the smaller scaling factor of 0.475 is selected for scaling to change the original picture size to 608×456. Next, 76 black regions are filled above and below the picture to meet the input requirement of 608×608. Subsequently, through the convolutional layer and stride, downsampling is set to gradually reduce the spatial resolution of the feature map while increasing the channels of the feature map. Multiple convolutional layers are used to extract features and output the feature map. Batch Normalizaton is added after the convolutional layer to accelerate training and improve the stability of the model. Next, the output feature map is divided into two parts by CSP. One part undergoes further convolutional operations, and the other part is directly passed to the next layer. Finally, the processed features and the unprocessed features are merged to form a new feature map. The features of the ladle, ladle trunnion, and hook in the new feature map are further identified. For the recognized hook rectangular box, the distance between the midpoints of the diagonals of the hook rectangular box and the ladle trunnion rectangular box is taken. If the distance between the two points is greater than a specific value, it is determined that the current is not firmly hooked.

5. The overhead crane safety monitoring system based on YOLO V5 and Wincc according to claim 1, characterized in that, Data acquisition module, image vision module, lifting state judgment module, and abnormal alarm module achieve data exchange with the overhead crane PLC through a 5G CPE device using the OPC protocol.

6. The overhead crane safety monitoring system based on YOLO V5 and Wincc according to claim 1, characterized in that, In the lifting state judgment module, to further confirm the judgment result, the ladle weight is introduced as an auxiliary judgment means.

Images