Automatic Driving Tracking Algorithm Based on DeepSort Tracking Algorithm

Abstract

Based on DeepSort algorithm, this paper uses identity mapping technology to map the location of historical identity to the current feature map, which greatly improves the tracking accuracy.In addition, in order to ensure real-time tracking, the self-attention module in Transformer is linearized, which reduces the time complexity of the algorithm and greatly improves the algorithm speed.Finally, the improved algorithm is tested to verify the feasibility and practicability of the algorithm.