The Visual Object Tracking Algorithm With Sparse Representation

Visual object tracking plays a key role in numerous vision applications including security surveillance,human-computer interaction,activity recognition and so on.How to effectively model the appearance of the target and how to accurately infer the state from all candidates are two key steps for a successful tracking system.While a lot of object tracking methods and much progress has been made in the past decades,developing a robust tracker is still a challenging problem.The main difficulty of visual tracking is designing an effective appearance model which should not only distinguish the target from the background but also be robust to its inevitable appearance variations of the target object,which include intrinsic and extrinsic factors in a dynamic scene.Therefore,it is still a challenging task to design a robust visual object tracking algorithm.Recently,researchers have introduced sparse representation for visual tracking and it is solved through a series of L1 minimization problems to solve the model tracking problem.Motivate by this work,in this paper,we propose a robust object tracking algorithm with a structured sparse representation model.This model includes one fixed template,nine variational templates and the background templates,which are selectively updated to adapt to the appearance change of the target.By incorporating the block-division feature into sparse representation framework,it can capture the intrinsic structured distribution of sparse coefficients effectively and is more conducive to handle occlusions and other challenging problem.Then,we propose a sparse representation-based discriminative classifier to separate the target from the background.Furthermore,we adopt the incremental principal component analysis(PCA)method that exploits the strength of both subspace learning and sparse representation for modeling the updated template.Both qualitative and quantitative evaluations on benchmark challenging sequences demonstrate that the proposed tracking algorithm performs favorably against several state-of-the-art tracking methods.Meanwhile,in this paper,we propose a robust visual tracking algorithm in a cotraining framework.The proposed object appearance model exploits the strength of both holistic representation and local histogram.The proposed tracking method is effective in dealing with appearance changes through incremental subspace learning and the computation complexity is reduced.Experimental results on several challenging sequences demonstrate the robustness and effectiveness of the proposed algorithm,especially when the objects exhibit large appearance changes.