Correlation Filter-Based Visual Tracking with Multi-featured Adaptive Online Learning

Abstract

While discriminative correlation filter (DCF) has attracted much attention due to its excellent computational efficiency and robustness, factors like occlusion, motion blur, deformation and background interference cause tracking failure. To address these issues, this work proposes improved background-aware correlation filter (BACF) that utilize colour features as a complement of Histogram of Oriented Gradient (HOG) to improve the representation of the target, and implementing adaptive feature fusion in the response layer using Peak to Sidelobe Ratio (PSR) as a reference metric. Moreover, to further reduce the risk of model drift, dynamically adjusting the learning rate to adapt to complex background. Extensive experiments on the OTB-50, OTB-100 benchmarks have shown that the proposed performs well compared against many state-of-the-art trackers, achieving an AUC score of 79.6% on the OTB-100.