Research On Vehicle Flow And Flow Direction At Road Intersection Based On Video And Application Of Signal Timing

With the rapid development of economy,the increasing number of motor vehicles has brought tremendous pressure to road traffic.The most serious problem is urban road traffic congestion.As the node of urban road traffic,intersection is one of the important parts of urban road traffic system.Therefore,solving the problem of intersection congestion is of great significance to alleviate the problem of urban road traffic congestion.Obtaining the traffic volume of Road intersection and making reasonable signal timing is an effective means to solve the problem of intersection congestion.In this paper,a new framework of traffic statistics based on three stages is proposed,that is,vehicle detection by improving YOLOv3 target detection model,improved Hungarian Calman filtering algorithm for target tracking is used to track the detected vehicle,and on the basis of vehicle detection and vehicle tracking,different virtual detection lines are set up according to different lane types,so as to realize traffic volume statistics of tracking vehicles in each entry lane of intersection.Firstly,aiming at the randomness of initial cluster center selection in YOLOv3 algorithm,K-means++ clustering algorithm is used to improve it,and the network structure and loss function of YOLOv3 algorithm are fine-tuned and improved.The experimental results showthat the mAP of the improved YOLOv3 algorithm has been improved to a certain extent;Secondly,for the monitoring video of intersections,the problem of tracking vehicle loss is caused by mutual occlusion between vehicles.In the Hungarian Kalman filter algorithm,two strategies are adopted to improve the Hungarian Kalman filter algorithm according to the length of tracking vehicle loss time,which enables the vehicle to be tracked effectively even under occlusion.Finally,this paper proposes a traffic statistics method for different lanes,that is to say,different virtual detection lines are set up according to different lanes,and traffic statistics are carried out for tracking vehicles with different directions in the inlet lanes.This paper selects the video surveillance of Fuzheng intersection on Qingrong Road,and carries out the statistics of the maximum 15 minutes traffic volume in the peak period of 5 days.The statistical results show that the traffic statistics method has higher accuracy and feasibility.Then the traffic volume of the intersection is input into the signal timing software.According to the design flow of signal timing,the optimal signal timing scheme is worked out,which improves the traffic efficiency of the intersection and provides an important basis for the traffic management department to make reasonable traffic decisions.