Research on highway night traffic event detection method based on video processing

At present, most highways in China have established full-coverage video systems, but highway traffic event detection still uses alarms and manual inspections. Especially, the high incidence of nighttime freeway traffic events makes inspection difficult, which is a difficult point in freeway emergency management. We develop a nighttime traffic event detection algorithm that can be integrated into existing highway video surveillance systems, which has three components: video initial processing, vehicle detection and vehicle tracking speed measurement. Among them, the Hom-Schunck optical flow algorithm is used for vehicle detection and Deepsort’s multi-target vehicle tracking algorithm is used for vehicle tracking and speed measurement. Finally comparing the vehicle detection part using background difference method, yolo algorithm and the algorithm in this paper, the algorithm in this paper has higher detection accuracy, while the detection speed meets the system working requirements.