Learning residue-aware correlation filters and refining scale for real-time UAV tracking

Abstract

Unmanned aerial vehicle (UAV)-based tracking finds its applications in agriculture, aviation, navigation, transportation and public security, etc and develops rapidly recently. However, due to limitations of computing resources, battery capacity, requirement of low power and maximum load of UAV, the deployment of deep learning-based tracking algorithms in UAV is currently not feasible and therefore discriminative correlation filters (DCF)-based trackers have stood out in UAV tracking community for their high efficiency and appealing robustness on a single CPU. But confronted with difficult challenges the efficiency and accuracy of existing DCF-based approaches is still not satisfying. Inspired by the good optimization properties associated with residue representation, in this paper we exploit the residue nature inherent to videos and propose residue-aware correlation filters which demonstrate better convergence properties in filter learning. In addition, we propose a scale refinement strategy to improve the wildly adopted discriminative scale estimation in DCF-based trackers, which, in fact, greatly impacts the precision and accuracy of the trackers since accumulated scale error degrades the appearance model as online updating goes on. Extensive experiments are conducted on four UAV benchmarks, namely, UAV123@10fps, DTB70, UAVDT and Vistrone2018 (VisDrone2018-test-dev). The results show that our method achieves state-of-the-art performance in UAV tracking.