Object Tracking with Soft Discriminative Sparse Dictionary

In this paper,a soft discriminative tracking method with adaptive parameters is proposed which is based on online sparse template.Firstly,an off-line local-constrained dictionary based on HOG is used to represent the object,which can make this tracking algorithm more robust.Then a dynamic parameter observation model is constructed with online sparse dictionary and the linear classifier which is learned from the local-constrained dictionary.In the end,particle filter is used to estimate the states of the object.The experimental results on several challenging image sequences demonstrate that the proposed tracker achieves more favorable performance than the state-of-art methods.