Research on Aircraft Tracking Technology Based on Improved DeepSORT Algorithm
Yanwen Zhang, Miao Wang, Yuwen Fu · 2023
The aim of this paper is to explore aircraft tracking techniques based on the improved DeepSORT algorithm to solve practical problems such as small aircraft targets at long distances and in complex weather environments. This paper firstly reviews the classical algorithms and techniques of target tracking, and focuses on the DeepSORT algorithm and its application in aircraft tracking. On this basis, this paper improves the DeepSORT algorithm, including the optimisation in terms of introducing the attention mechanism in the target detection part, adopting the bi-directional long and short-term memory network model in the trajectory prediction part, and introducing the residual connection and batch normalisation techniques in the feature extraction part. In this paper, the improved DeepSORT algorithm is comprehensively evaluated and validated by conducting a large number of experiments on the public dataset UAV123. The experimental results show that the improved algorithm in this paper achieves significant performance enhancement in the vehicle tracking task, with an improvement of about 4.7 in accuracy, about 5 in FPS, and about 2.6 and 1.4 in MOTP and MOTA, respectively, and the study provides useful references and lessons for the further development of vehicle tracking technology.