Research Of Marginal Distribution Bayesian Filter In Multiple Maneuvering Targets Tracking

The task of multiple maneuvering targets tracking is to estimate the number and state of maneuvering targets in the present of noise and clutter. Tracking multiple maneuvering targets and estimating the time-varying number of targets are the focus and difficulty of current research in the academia. The traditional multiple maneuvering targets tracking algorithm is restricted in practical applications because it demands data association. The multi-target Bayesian filter which is based on finite set statistics(FISST) avoids data association, hence attracting substantial interests from many researchers. The probability hypothesis density(PHD) filter is an approximation implementation of the multi-target Bayesian filter. However, the PHD filter cannot distinguish multiple targets when these targets are closely spaced; therefore it is restricted in practical applications of multiple maneuvering targets tracking. Aiming at this problem, the marginal distribution Bayesian filter for multiple maneuvering targets is proposed in this dissertation, which is summarized as follows:1) In chapter 2, the FISST theory, the multi-target Bayesian filter based on FISST, and the PHD filter are introduced. The FISST theory provides a mathematical foundation for multi-target tracking algorithms. The multi-target Bayesian filter based on FISST propagates the joint distributions of the multi-target state in the filter recursion. The PHD filter is an approximation implementation of the multi-target Bayesian filter. It propagates the first-order moment of the joint distribution in the filter recursion.2) In chapter 3, we study the marginal distribution Bayesian(MDB) filter. Based on this filter, an implementation of marginal distribution Bayesian filter in linear and Gaussian system is developed. The MDB filter uses the existence probabilities to express the uncertainties of individual targets, and it propagates the marginal distributions and existence probabilities of each target in the filter recursion. The simulation results demonstrate that the MDB filter has a good tracking performance in the presence of clutter as well as target appearance and disappearance.3) To solve the problem of multiple maneuvering targets tracking in linear and Gaussian systems, we combine the interacting multiple models and the jump Markov system models with MDB filter to develop two implementations of MDB filter for tracking multiple maneuvering targets. The first is the interacting multiple model marginal distribution Bayesian filter, and the second is the jump Markov system model marginal distribution Bayesian filter. The simulation results demonstrate that both two filters can track multiple maneuvering targets stably and accurately in the presence of noise and clutter. In addition, we also analyze the application scope of two proposed filter through simulation experiments.