Visual object tracking via multi-view and group sparse representation
Borui Mo, Ke He, Aidong Men · 2016
The use of multiple features for tracking has been proved as an effective approach, for the reason that limitation of each feature could be compensated. Sparse representation with a particle filter is one of the most influential frameworks for visual tracking. Many sparse representation tracking algorithms consider the holistic or local representation of the target, and rarely take into account distribution of sparse coefficient. In this paper, we propose a group sparse representation model using multiple features for object tracking. By appropriately selecting features to form the dictionary, the group sparse based representation scheme can assign proper weights to each group of selected templates. As a result, the potential target in each frame can be identified precisely. Experimental results on public available videos demonstrate that the proposed method performs favorably against several state-of-the-art tracking methods.