Visual Tracking Based On Random Projection And Sparse Representation

Object tracking is important in intelligent video processing,which is widely used in many areas,such as industry production,military defense and safety surveillance.As tracking problems are various and complex,many difficulties should be solved in tracking study.A robust object tracking method is proposed to deal with technical problems during tracking.The algorithm can be divided into four parts: the appearance model,pictures' dimension reduction,the motion model and template update scheme.Firstly,the appearance model is built by the global discriminative classifier based on sparse representation.To separate object from background,positive and negative templates are built in first frame to represent foreground object and background.Secondly,random projection(RP)is used to reduce dimension of templates and candidate objects,which could release calculation burden of 1L problem and improve tracking efficiency.Thirdly,particle filter(PF)is used to estimate the object motion,and the local weighted posterior probability method is proposed to compute the posterior probability,furthermore,the multi-normal resample method is used to maintain the diversity of particles.Finally,template update scheme is designed to alleviate module drift problem.The positive templates are divided into static templates and changeable templates,while different templates should be dealt with different way,and sparse reconstruction error is used to decide whether the object is occluded.Experiment results on numerous challenging videos show that the proposed method has better performance in accuracy and stability,in comparison with state-of-the-art tracking methods.