Application Of Particle Filter In Target Tracking

Recently, The Particle Filter, which has been proved to be highly suitable for nonlinear and non-Gaussian problems, becomes the dominant technique for the target state tracking. The thesis starts with discussing the principles of estimation for nonlinear discrete-time systems, from Kalman filter to Particle filter, and then focuses on developing a novel algorithm to improve the particle filter for reentry vehicle tracking. Compared with other nonlinear filters, a better performance can be achieved. Another topic of the thesis is data association in multiple target tracking. The development and increasing sophistication of tracking systems, both civilian and military, has generated interest in algorithms of tracking large number of objects. The high bandwidth and sensitivity of modern sensors can lead to huge data loads and the presence of countermeasures can make tracking very difficult. The crux of the multiple target problem is to carry out data association process for measurement whose origin is uncertain due to random false alarms, clutter, interfering targets or other countermeasures. The most common algorithm for data association is the joint probabilistic data association filter. Therefore, this thesis presents one target estimation in the presence of clutter using probabilistic data association with extended Kalman filter and Particle filter, and two target tracking by using joint probabilistic data association with extended Kalman filter. Simulation results show good performances for multiple target tracking.