Research On Video Surveillance Technology Of Unguarded Railway Crossing Based On YOLO

With the gradual development of my country’s railways,there are more and more motor vehicles passing unguarded railway crossings,and there may be uncertain factors such as people and other vehicles passing through the crossing during the normal operation of the railway system.Therefore,it is urgent to pay attention to safety issues in the management of unguarded railway crossings.However,the existing video monitoring devices for unguarded railway crossings at this stage are not highly intelligent and have obvious defects.Therefore,intelligent safety detection for unguarded railway crossings has become a key research direction for traffic researchers.In recent years,with the popularization and application of unattended railway crossing surveillance cameras,automatic detection of targets through image processing technology has also become a research hotspot.The thesis sorts out the detection and processing flow of video surveillance technology for unguarded railway crossings.Based on the existing unguarded railway crossing surveillance video device,the video detection system process is designed and the code of the video detection technology is developed.,And finally realized the safety detection and video detection and early warning functions of unguarded railway crossings.The paper analyzes and studies the algorithm structure and process of video detection technology,and then uses the traditional inter-frame difference method to detect moving vehicles and pedestrians in the video images of crossings,and uses Snake algorithm to track moving targets.However,this algorithm can only basically meet the needs of target detection,cannot distinguish between vehicles and pedestrians,and has a low detection accuracy rate,which cannot meet the requirements of video surveillance technology for unguarded railway crossings.Therefore,the paper studies the characteristics of moving targets and the condition of unguarded railway crossings,compares the pros and cons of each generation of YOLO versions,and selects a target detection algorithm based on the YOLO-v3 algorithm to realize the detection and tracking of moving targets,and the algorithm It can distinguish between people and vehicles and detect the target speed.The alarm function can be realized when an imminent safety hazard is detected.In the video detection system based on YOLO-v3,this paper uses the Kalman filter tracking algorithm,and completes the target detection training in the COCO data set,the tracking training in the MARS data set,and the video detection system completes the detection of moving targets,Tracking and early warning.This paper studies the test video collected from real scenes of unguarded railway crossings.The experimental algorithm uses python and other software as the programming platform.The research results show that the unguarded railway crossing video surveillance technology based on YOLO can complete the monitoring of vehicles and pedestrians in the video.The detection accuracy of moving targets is high,and the accuracy of moving target detection is high,up to 97.3%,which can provide reference for the safety management of unguarded railway crossings.