A Roadside Vehicle Tracking Scheme based on Predicting Vehicle Area Occlusion

Roadside vehicle tracking is generally used in outdoor intelligent parking scenarios. Using elevated cameras, track vehicles entering and exiting roadside parking spaces in the open air and time them to calculate parking fees. In this scenario, moving vehicles will approach stationary vehicles parked in the parking space, which is prone to id switch phenomenon. To reduce the frequency of id switches, I propose a new tracking scheme called OcclusionTrack. The scheme first divides the vehicles into four areas and predicts whether each area is obstructed by other vehicles, and adjusts the size of the tracking box accordingly to improve tracking accuracy. After a large number of comparative experiments, the method proposed in this article has obtained competitive results.