Research On Tower Crane Safety Monitoring System Based On Improved Yolov4

Tower crane is an indispensable auxiliary production tool for construction projects,and its use is increasing with the expansion of the scale of the construction industry.Due to the structural characteristics and illegal operation of tower cranes,it is easy to lead to tower crane safety accidents.In order to ensure the life safety of tower crane drivers and meet the modernization requirements of tower crane operation safety,intelligence and real-time,this paper designs a tower crane safety monitoring system through the study of hook tracking method,and the main work is as follows:(1)In view of the shortcomings of the existing tower crane safety monitoring system with scattered functions and lagging monitoring screens,a scheme is proposed to connect the camera,data acquisition and transmission module and the host computer into a local area network through routers,switches,wireless bridges and other equipment,so that the tower crane on-site monitoring screen and monitoring data are jointly displayed on the web page of the intranet to ensure the real-time transmission of video data.(2)In order to improve the speed and accuracy of hook detection,a hook detection algorithm based on improved YOLO4 is proposed.The original network model is used to detect hooks in the tower hook dataset,and the experimental results show that YOLOv4 can achieve the best balance of hook detection speed and accuracy.The improved YOLOv4 algorithm is used for experiments and analysis,and the results show that the average accuracy(MAP)of the optimized algorithm is increased by 1.9%,and the detection speed(FPS)is accelerated by 3frames per second.(3)Design the software and hardware of the tower crane safety monitoring system.The system hardware part is the data acquisition and transmission module,which mainly includes the microprocessor control module,the data acquisition module,and the serial port to Ethernet module.The system software part is mainly the host computer monitoring system,and its data comes from two parts,one part is the tower crane operation data sent by the data acquisition and transmission module to the local area network,and the upper computer monitoring system receives and analyzes,displays and stores the data and abnormality warning.The other part is the video data transmitted by the camera to the LAN in real time,and the host computer monitoring system integrates the camera on the web end to display the on-site monitoring video of the tower crane.At the same time,in order to reflect the application value of the algorithm and solve the problem of difficult docking between the algorithm and the Java development system,the C++ training weight file is used to realize the docking of the C++ training model and the Java system through Java Open CV,and the hook monitoring video stream is pulled into a picture for the improved YOLO algorithm to process,and the ROI value output by the algorithm is converted into the PTZ parameter of the camera to realize the intelligent tracking of the hook.(4)In view of the problem that the system early warning module will be paralyzed when the hardware device fails in the traditional hardware alarm mode,the software early warning method is added to ensure the reliability of the system alarm when abnormal data is detected.However,the real-time warning of software mode is worse than that of hardware alarm,so big data technology is used to ensure the real-time performance of the system.That is,the Netty server serializes the received and parsed data and transmits it to the Kafka queue for the Flink program to listen,and the Flink program compares the real-time data with the threshold data in the database,and once the data is detected,the abnormal warning is immediately given.