Application Of Levy Distribution In Particle Filter

Video target tracking involves all aspects of life;particle filtering is a classic algorithm used for target tracking and is widely used to achieve target tracking in video images.Compared with Kalman filter algorithm,particle filtering can achieve good results in nonlinear and non-Gaussian systems.However,since the resampling stage of particle filtering will largely replicate particles that with larger weight,this will directly discard the information of particles that with smaller weight,thus affecting the accuracy of the particle filter algorithm's estimation of the real state.Taking into account particle depletion,this paper proposes to use Levi Flight to optimize the particle filter algorithm in combination with Levi Flight's operation mechanism.This algorithm combines the uniformity of Levi's flight and the great exploration ability in the unknown region with particle filter to improve the accuracy of particle filtering algorithm,The main algorithm improvements and results of the paper are as follows:First,Levy-PF algorithm that combines Levi Flight and particle filtering is proposed.Combining the search ability of Levi Flight to the unknown space and particle filtering that have greatly improved the particle depletion caused by copying a large number of weights in the resampling phase of the particle filter.Second,test the accuracy of the improved algorithm in two dimensions space.By comparing the standard particle filter algorithm with the particle filter algorithm optimized using particle swarm optimization,the three algorithms compare the error through the state model and the test data of the observation model.Third,test the diversity of particles that are in the video target tracking application.In the experiment,we used the standard particle filter algorithm and Levy-PF algorithm to compare particle diversity after particle resampling to illustrate that the proposed algorithm has advantages in global search capability and particle diversity.Through the comparison of filter accuracy test,particle diversity test,and particle diversity in the video target tracking,Levy-PF algorithm proposed in this paper shows that the algorithm is superior to the original algorithm in terms of filter accuracy and particle diversity.It has some practical significance,and the Levy-PF can also improve the ability of global search.