Bilateral Weighted Regression Ranking Model With Spatial-Temporal Correlation Filter for Visual Tracking

Many discriminative correlation filter (DCF)-based methods have successfully leveraged the guidance for solving two problems (i.e., boundary effect and temporal filtering degradation) as a model prior to visual tracking. The intuitive motivation of these is control degeneration updating loss objective function with structural framework. While rely mostly on various regularization items, they always ignore from data fidelity term. Therefore, we propose bilateral weighted regression ranking termed BWRR. Here, resort procedures above problems. First, BWRR introduces constraint into term rows columns learning matrices could impose an adaptive penalty large during process avoid degradation problem. Second, updated not directly applied calculation each iteration. Instead, new product matrix obtained by numerical transformation filter. We show that proposed converts original problem regression-with-ranking problem, thus avoiding positive negative sample imbalance. Overall, iteratively solved alternating direction method multipliers(ADMM). Qualitative quantitative evaluations demonstrate effectiveness superiority our extensive experiments OTB, VOT, UAV datasets.