ARTracker: Compute a More Accurate and Robust Correlation Filter for UAV Tracking

Unmanned aerial vehicle (UAV) tracking focus on tracking moving targets from flying platforms, where the target undergoes a lot of aspect ratio changes, that is, viewpoint change and rotation. The discriminative correlation filter (DCF)-based method shows a promising solution to UAV tracking due to its high computational efficiency. DCF-based trackers apply a fixed-bandwidth Gaussian function label for model training and incremental update to adapt to the dramatic appearance changes in the target during tracking. However, due to the poor regression ability of the correlation filter, DCF trackers are unable to describe the target state when the aspect ratio changes, thus hurting the tracking performance. To alleviate this, we propose a novelty correlation filter, which constructs a Gaussian-like (GL) function label for correlation filter training. The label fully considers the aspect ratio distribution of the target, which facilitates the training of a more robust tracker. Furthermore, an accurate incremental update is proposed to mitigate model degradation by combining target samples with adaptive aspect ratios. Extensive experiments are conducted on three popular UAV benchmarks, that is, VisDrone2018-test-dev, UAV20L, and DTB70. Results well demonstrate the superiority of the proposed method over both DCF and deep-based trackers.