Particle Filter Methods In Radar Multi-Target Tracking

With the rapid development of China's shipping,the environment of shipping is becoming more and more complex,and the future tracking system is facing severe challenges.Multi-target tracking has been regarded as a research focus and difficulty by scholars,it mainly includes tracking correlation and tracking filtering,a good tracking filtering algorithm can improve the tracking accuracy greatly.Compared with the general filtering methods,particle filter has better robustness and unique advantages in dealing with non-linear and non-gaussian problems.The dissertation focus on the research of particle filter algorithm in multi-target tracking.Firstly,the basic concepts and principles of multi-target tracking and particle filtering are analyzed,the Bayesian theory and Monte Carlo integration are introduced.The data association and tracking filtering theory in multi-target tracking are mainly discussed,and the commonly used methods of data association and tracking filtering are compared.Secondly,the Sequential Importance Resampling Particle Filter algorithm and Cost Reference Particle Filter algorithm are studied,and then a Gauss Cost-reference Particle Filter algorithm is proposed,called GCRPF.A gaussian distribution is used in the algorithm to replace the resampling process based on the Cost Reference Particle Filter algorithm,it combines the advantages of the Cost Reference Particle Filter andt greatly reduces the complexity of the algorithm,so the operation speed of the algorithmit is improved.And the simulation is carried out to demonstrate that the improved algorithm reduces the amount of calculation and shortens the running time on the premise of guaranteeing the filtering effect.Finally,the Joint Probability Data Association algorithm in multi-target tracking system is studied.Based on MC-JPDA,a Cost-Reference Particle Filter-Joint Probabilistic Data Association is proposed,called CRPF-JPDA.The algorithm does not need to know the statistical characteristics of the system,and for each of the targets,the connectivity probabilities are approximated by the particles with different costs.CRPF is utilized for approximated the prediction and update distributions.And the simulation is carried out to demonstrate that the CRPF-JPDA algorithm realizes multi-target tracking in the case of unknown system statistics,and the the tracking precision is improved.