Particle Filter Algorithm Application And Research On Target Tracking

The target tracking is the problem of a research exercise on the tracking target that it can'taccurate the description of target sport estimate. The subject has the extensive application inthe military and civilian realms. Firstly, this dissertation summarizes the establishment of thetarget tracking models, and gives some of familiar target maneuvering models and introducesthe method of sitting the mark conversion for measurement model. Sequentially, we researchsome important filtering methods that include Kalman filter, extended Kalman filter,unscented Kalman filter and particle filter in the target tracking areas. Then aiming at thenonlinear and/or non-Gaussian filter problems, the generic ideas of particle filter are given.Developing in 90's last century, Particle filter, a new filter method based on Monte Carloand recursive Bayesian estimation, has special advantages in dealing with the state and theparameter estimation in the nonlinear and non-Gaussian system. However, the disadvantagesof complex algorithm architecture, enormous computations and low speed have restricted itsimplementation in real-time system. In the dissertation the problems of particle filter arediscussed and some improvement methods are illustrated. Pseudo-code of every variants ofparticle filter is given. Several variants of particle filter such as SIR, ASIR are compared. Theadvantages and disadvantages of them are discussed.At last, though a simulation, particle filter is compared with extended Kalman filter andunscented Kalman filter. It is known that target tracking method based on extended Kalmanfilter performs well when tracking a target in a linear system or with small maneuver but badwhen in a strongly nonlinear system or with high maneuver. In a simulation comparison withextended Kalman filter and unscented Kalman filter, it is proved particle filter betterperformance when in a strongly nonlinear system or with high maneuver.