Research On Target Tracking Algorithm Based On Sparse Representation And Scale Invariance In Polar Coordinates

Target tracking is one of the research hotspots of computer vision,which has wide practical applications in such fields as entertainment,medical treatment,transportation,finance and military defense.The development history of target tracking shows that the target tracking technology has been improving.However,it is still a huge challenge to establish a robust target tracking system due to the influence of internal and external factors,such as illumination change,deformation change,rotation and occlusion.In recent years,the target tracking algorithm based on the generated model and the discriminant model have drawn wide attention from researchers in universities and scientific research institutions due to their respective advantages.In this paper,particle filter target tracking algorithm based on generation model and correlation filter target tracking algorithm based on discrimination model are studied.The main research works are as follows:Most traditional sparse representation target tracking algorithm adopt only a single appearance model,which may cause tracking failure in the context of existing similar object,partial occlusion,illumination and/or pose variation,and rotation,et al.We propose a target tracking method based on multiple appearance model fusion with structured sparse representation in particle filter framework.In the proposed method,the global and local features of the target are sparsely represented respectively by global and local templates,the spatial structure information of the target in the local appearance model is characterized by summation of the sparsely coding coefficients of each local patch of one candidate target on local target templates that have the same position as that of the local patch.Meanwhile,the global and local templates are updated respectively based on sparse representation as well as the incremental principal component analysis along with block detection.The proposed template update method can reduce computational complexity and improve the efficiency of the appearance model update.Experimental results on video sequence dataset show that the proposed method outperforms existing popular tracking algorithms in the aspect ofrobustness and accuracy,The proposed target tracking algorithm can track target robustly in the context of existing similar object,partial occlusion,illumination or pose variation,and rotation.To improve the target scale adaptive performance of correlation filter trackers,a method of target scale invariance in polar coordinates is proposed.In this method,multiple sets of different features are used to learn the correlation filter,and then phase correlation theory is used to address the target scale change problem in log-polar coordinates.The experimental results show that,compared with the current advanced visual target tracking methods,the tracking method proposed in this paper can track the target accurately and efficiently when the target has a large scale change.It is shown that the proposed method can maintain the efficiency and simplicity of the traditional discriminant correlation filter tracking method and has good adaptive ability at the target scale.