Application Of Sparse Representation In Target Tracking

With the development of science and technology, the research of computer vision has become more and more deeply. There are more and more practical applications in computer vision. It has a great influence on people's daily life and production. In the research of computer vision, the research of target tracking has made great progress.A large number of target tracking algorithm based on sparse representation are summarized and in-depth study, according to the challenge problems in the field of target tracking, a target tracking algorithm based on sparse representation combined appearance model is proposed. In the generative model, we use alignment pooling on obtained the block structure of the sparse coding coefficients. Then sparse coefficients are combined with weighted processing, and thus preserve the spatial structure of the target and local information to improve the robustness of target positioning precision and tracking. In complex background, in order to better separate the target and background, the judgment model of sparse representation is used to enhance the accuracy of tracking. Finally, a more robust joint appearance model is obtained.In order to cope with the change of target appearance and a series of difficulties in the process of tracking, the dictionary template is updated in real time. The based on sparse representation and incremental principal component analysis method of dictionary update method, and in selected need to be updated dictionary problem, this paper proposes a reverse sparse representation method was used to calculate the weight of the template in the dictionary, both to preserve the target of the original appearance, but also in order to reflect the current goal of morphology, the update of the dictionary is more accurate.In the motion model, particle filter is used to achieve the target tracking. To improve the traditional motion model of particle filter, the reconstruction error is used to choose particle. Experimental results show that this method guarantees the accuracy of tracking while reducing the complexity of the algorithm.The proposed algorithm is compared with many mainstream algorithm in a large number of test videos, Experiments show that the proposed algorithm has higher accuracy and robustness.