Target Tracking Based On Sparse Presentation

It is a great challenge for intelligent analysis and discovery of useful information from a great massive video resources,and the target tracking become more and more popular as the important part of the intelligent video surveillance.The feature extracted by the traditional method cannot keep the invariance of the target when the target surfers occlusion,illumination and pose variation in the moving process,which leads to the failure of the tracker.To handle the problems mentioned above,the paper conduct a survey on the target tracking based on sparse representation.After a systematic study and summary of sparse representation,the paper constructs a target tracking framework based on sparse representation and particle filtering.A structured object appearance model based on sparse representation is proposed.To handle the effect of occlusion on template updating,the target subspace is obtained by incremental learning.On the basis of this,a new method for online template updating based on subspace representation is implemented.In view of the problem that a large amount of time is needed to solve the problem of sparse expression optimization,two kinds of methods are proposed,which are based on the observation likelihood upper bound and the time and space.The effectiveness of the proposed algorithm is verified by experiments on a number of standard test sets.The average center error and the success rate of the target tracking are similar to the classical algorithm,while the time costed is 3 times more than that.