A Kind of Cross-Camera Tracking Strategy in Tunnel Environment

In view of the fact that the traditional single camera tracking method cannot meet the real-time changing needs in the specific expressway tunnel environment, a Multi-Target Multi-Camera Tracking algorithm is proposed in this paper. Firstly, the target detection model based on PP-PicoDet is trained, and the accuracy of the risk object detection is improved by using the static object combined filtering method. Secondly, an improved DeepSORT algorithm is used to track the object vehicles in real-time. Then a Lane Based Time Mask algorithm is designed to constrain the track matching domain according to the characteristics of the tunnel. Finally, a Local Camera Group Clustering method based on hierarchical clustering is used for vehicle trajectory fusion under different cameras. The experimental results show that the algorithm realizes the real-time tracking and risk warning in the expressway tunnels with high precision.