Robust Object Tracking via Sparse Collaborative Appearance Model

Abstract — In this paper, we propose a robust object tracking algorithm based on a sparse collaborative model that exploits both holistic templates and local representations to account for drastic appearance changes. Within the proposed collaborative appearance model, we develop a sparse discriminative classifier (SDC) and sparse generative model (SGM) for object tracking. In the SDC module, we present a classifier that separates the foreground object from the background based on holistic templates. In the SGM module, we propose a histogram-based method that takes the spatial information of each local patch into consideration. The update scheme considers both the most recent observations and original templates, thereby enabling the pro-posed algorithm to deal with appearance changes effectively and alleviate the tracking drift problem. Numerous experiments on various challenging videos demonstrate that the proposed tracker performs favorably against several state-of-the-art algorithms. Index Terms — Object tracking, collaborative model, sparse representation, feature selection, occlusion handling