Adaptive Channel Selection for Robust Visual Object Tracking with Discriminative Correlation Filters

Abstract Discriminative Correlation Filters (DCF) have been shown to achieve impressive performance in visual object tracking. However, existing DCF-based trackers rely heavily on learning regularised appearance models from invariant image feature representations. To further improve the of DCF accuracy and provide a parsimonious model attribute perspective, we propose gauge relevance multi-channel features for purpose channel selection. This is achieved by assessing information conveyed each as group, using an adaptive group elastic net inducing independent sparsity temporal smoothness solution. The robustness stability learned are significantly enhanced proposed method process selection performs implicit spatial regularisation. We use augmented Lagrangian optimise discriminative filters efficiently. experimental results obtained number well-known benchmarking datasets demonstrate effectiveness method. A superior over state-of-the-art less than $$10\%$$ 10 % deep channels.