Target detection and tracking of traffic flow at intersections with camera
Minjie Du, Gang Tao, Weibin Zhang · 2022 34th Chinese Control and Decision Conference (CCDC) · 2022
This paper mainly studies the extraction of traffic flow parameters from intersection surveillance videos, which is an important issue of traffic management. The fusion algorithm of YOLO and DSST (Discriminative Scale Space Tracker) is used to obtain time headway and PCU (Passenger Car Unit) at intersections. By narrowing the recognition range and generating hotspots during target tracking, the recognition effect is enhanced. Experiments show that the optimization algorithm can more accurately extract intersection traffic flow parameters. Compared with only using the recognition algorithm, the average PCU accuracy rate of the proposed algorithm in this scenario is increased by 6.8%. Compared with using the YOLO+DSST algorithm, the average correct rate of the proposed algorithm in this scene is increased by 3.15%.