Traffic flow statistics algorithm based on YOLOv3

Abstract

Real time traffic flow monitoring is not only of potential commercial value, but also one of the important means to ensure public traffic safety. Video recognition technology is the most effective way to realize traffic flow monitoring. In view of the current situation that the accuracy of existing traffic flow statistics algorithms is poor, and it is easy to cause false detection and false detection, a multi lane traffic flow statistics algorithm based on YOLOv3 is proposed. Firstly, the feature extraction network is used to extract features from the input image, and the image position and category probability value are predicted; secondly, the vehicle position detected in the adjacent two frames is compared, and whether the vehicle in the two frames is the same vehicle is judged according to whether the center point of the vehicle marker box in the adjacent two frames is at the same point, so as to achieve the purpose of tracking; finally, the set detection line position is used. The traffic flow in each lane can be obtained. This method can track the vehicles and count the traffic flow in any lane. The experimental results show that in vehicle tracking and traffic flow statistics, the problems of inaccurate detection and tracking caused by the adhesion of vehicle target area in traditional moving target detection algorithm and the limitation of virtual coil algorithm in multi lane traffic flow detection are solved, and the accuracy rate of traffic flow detection is 87.7%.