Target tracking for rotor UAV based on multi-scale compressive sensing

Abstract

This paper proposes a target tracking algorithm for the rotor unmanned aerial vehicle (UAV). Traditional compressive tracking (CT) is an efficient, fast and accurate method suitable for the real-time requirement of UAV target tracking. However, the tracking box size in it remains fixed, which is likely to bring errors quickly when the target has a big pose or scale variation. In this paper, we propose an improved multi-scale compressive tracking (IMSCT) algorithm combining compressive tracking and particle filter. IMSCT can adjust the tracking box size according to the target size variation or pose change, which decreases errors during the tracking process. The experiments show that it performs better than the original one and achieves good results in terms of accuracy, efficiency and robustness. The experiment outputs coordinates of the moving target in the image, which can be turned into yaw, pitch and roll instructions by the UAV flight control system to realize the real-time target tracking.