Research On Target Tracking Based On Particle Filtering

With the rapid development of control technology and computer technology, nonlinear filtering technique has found wide applications in many areas. The estimation accuracy of classical algorithms such as Kalman filtering and extended Kalman filtering decreases significantly and even appears divergence when applied to optimal estimation for nonlinear and non-Gaussian dynamic system. The particle filtering is a combination of Bayesian estimation theory and Monte Carlo method, which may apply to any nonlinear systems that can be expressed by state space model, as well as the nonlinear non-Gaussian systems that can not be expressed by the traditional Kalman filter, and has gained extensive application in target tracking fields. The accuracy of particle filtering can approximate to the optimal estimation.This thesis mainly investigates particle filtering method and its application in target tracking. The main work consists of the following aspects.1.Based on Kalman Filtering, the main nonlinear filtering methods including EKF, UKF and IEKF are analyzed and compared. In which the advantages and disadvantages of the algorithms are summarized.Simulations show that IEKF outperforms other algorithms.2.Considering that the selection of the importance density function has importantly affected on improving the particle degradation and filter accuracy, the IEKF is superior to EKF and UKF in the posterior distributions. This thesis combines particle filtering algorithm with IEKF.Simulation results show that the performance of the improved algorithm IEPF is superior to the EPF and UPF.In order to resolve the depletion problem after resampling,this thesis gives an improved algorithm through introducing a MCMC(Markov Chain Monte Carlo)step to increase the diversity of particles.The simulation results show that the accuracy of the particle filtering algorithm with MCMC is higher than that without MCMC,which can track the target more exactly.3.A new aigorithm PDA-IMM-PF combined the particle filtering algorithm with the probabilistic data association and interacting multiple model algorithms is given, and its application in target tracking is presented. Simulation results show that the PDA-IMM-PF algorithm with a higher accuracy is supweior to the PDA-IMM-UKF wich is based on UKF filtering algorithm.