A multi-objective traffic flow detection system based on an improved yolov4 algorithm

This paper uses deep learning related algorithms to solve the multi-target tracking and traffic statistics problems in the field of intelligent transportation. Firstly, the advantages of YOLOv4 algorithm are introduced, and then the YOLOv4 algorithm is used for field target detection, and selecting the statistical area to identify, track and count traffic targets in the video. Finally, comparing the detected traffic flow and the actual traffic flow results are analyzed, and then the improvement scheme of increasing Deepsort tracker is proposed for the problem that the algorithm produces repeated identification of the same target. After several rounds of testing, the combined algorithm can basically meet the needs of target tracking detection and traffic statistics in intelligent traffic scenarios.