Adaptive Maneuvering Target Tracking Based On Particle Filter

Target tracking is monitored to specific objectives, thus obtaining target position in the coordinate system, the motion parameters and trajectory.Maneuvering target tracking theory in areas such as national defense and civil has important application value.After the construction of the maneuvering target model, the need for the core part of the tracking system is tracking algorithm design. The traditional strong limitations Kalman filter algorithm, nonlinear system filtering problem is not resolved, then the proposed extended Kalman filter (EKF), also can not be applied in a number of non-linear systems.In recent years the development of particle filter (PF) is more and more attention of researchers in the field, due to the particle filter has low requirements for the environment.Then Julier et have proposed unscented Kalman filter (UKF) algorithm, which is higher than the EKF algorithm accuracy and significantly less than the calculated amount by the particle filter (PF) algorithm.But the main problem has troubled UKF algorithm,when is unknown noise statistics feature, UKF filter accuracy decline or even divergence.Scholars from various countries on the issues raised above various adaptive filtering algorithms, but the maximum likelihood method was more extensively studied. However, under normal circumstances, directly point system estimation of unknown parameters is very difficult,Maximum expected (EM) algorithm is to find the Maximum Likelihood or maximum a posteriori estimation algorithms probability model, using the EM algorithm, the noise the second moment of the calculation is simple, intuitive derivation, simplifying noise parameter estimation process.Work of this thesis is as follows:Firstly, in the PF algorithm based on maximum likelihood criterion, construct noise contain several statistical parameters of the likelihood function, and by expectation-maximization algorithm, the noise estimation problem into the log-likelihood function of mathematical expectation maximization problem, obtain PF algorithm based on maximum likelihood adaptive natural criteria and EM algorithm, then MATLAB simulation experiments to verify the algorithm, and finally get the estimated Gaussian noise case based Adaptive unknown process noise.Secondly,based on the above basis, to apply it to the case of non-Gaussian noise flicker more specific case, the combination of PF algorithm and EM algorithm process noise in the adaptive parameter estimation. While the proposed method in the flicker with examples UKF algorithm and EM algorithm is analyzed and compared MATLAB simulation comparison of the advantages and disadvantages of the two methods.Simulation results show that, adaptive UKF algorithm has been implemented, PF adaptive algorithm proposed new UKF algorithm effectively overcome non-Gaussian noise in the system in case the problem of declining filtering accuracy, and implement the statistical characteristics of the online system noise estimate.