Target Tacking Algorithm Research Based On Particle Filter

As one of the key technologies in the field of computer vision,target tracking plays an extremely important role in the fields of artificial intelligence,virtual reality,video surveillance,traffic detection,military,medicine and so on.At the same time,however,visual target tracking still faces many problems to be solved.In practical applications,targets are often interfered by various external factors,such as target scale changes,shape changes,rotation,partial or total occlusion and fast motions.In view of the above problems,both sparse representation and correlation filter are incorporated into the particle filter framework regarded as the core of our work to improve the tracking accuracy and robustness.The main research contents of this paper are as follows.(1)Target tracking algorithm based on an adaptive feature and particle filter is proposed.Firstly,all the particles are divided into two parts and put separately.The number of particles that are put for the first time is large enough to ensure that the number of the particles that can cover the target is as many as possible,and then the second part of the particles are put at the location of the particle with the highest similarity to the template in the particles that are first put,to improve the tracking accuracy.Secondly,in order to obtain a sparser solution,a novel minimization model for an L_p tracker is proposed.Finally,an adaptive multi-feature fusion strategy is proposed,to deal with more complex scenes.The algorithm can solve the problems of fast motion,deformation,illumination change and occlusion.(2)Target tracking algorithm based on particle filter and correlation filtering is proposed.Firstly,different search areas are separately set for the particle filter and the correlation filter to solve the problem of fast moving of the target,and the number of particles is reduced to improve the tracking efficiency.Secondly,combining with the properties of L_p norm and low rank,a new tracker minimization model is proposed to improve the accuracy and robustness of the algorithm.Finally,a novel model update strategy is proposed to solve the partial occlusion problem.(3)A reliable local target tracking algorithm based on particle filter is proposed.Firstly,a target partition method combining manual partitioning with random partitioning is proposed to make full use of the advantages of target patch,and a novel target location estimation strategy is proposed.Secondly,considering that the local target has the same motion law as the overall target,two new particl resampling rules are proposed to delete the particles that do not conform to the motion law to improve the tracking robustness by utilizing the motion trajectory of the particles to exploit the relationship between the particles.Finally,the target scale is estimated based on the change of the distance between the reliable particles,and an adaptive scale estimation method is proposed.In this paper,particle filter is used as the framework to improve the tracking accuracy and the robustness to target fast motion,deformation,illumination change and occlusion by mining the relationship between particles,combining correlation filtering and merging various features.