Research For Probability Hypothesis Density Filtering With Glint Niose

Along with the development of scient ific research and technology, target tracking method in the field of military and civilian have been widely used, and the probabilit y hypothesis densit y(PHD) filter is one of the hotest research in target tracking field in recent years. The PHD filter is filtering method based on finite set statistics(FISST). This paper mainly studies the PHD filtering method o f target tracking in glint noise environment under the framework of random finite set theory.In this paper, Chapter 2 firstly studies the characteristics of glint noise, the glint noise is in radar tracking environment, according to the significant characteristics of non Gaussian distribution and long tailed, consider using the t distribution modeling the approach. Then based on the theory of random finite set of the PHD filter algorithm, and considering the nonlinear observation equation in simulation experiments, the Gaussian mixture PHD(GM-PHD) impled by the extended Kalman filter(EKF) filter algorithm. By using the Student’s t-distribution to describe the glint noise statistics, the PHD filter is extended via augmenting the target state and the noise parameters.Chapter 3 studies by augment ing the target state and noise parameters of extended PHD filter. To derive a closed-form expression for the extended PHD filter, the prior Gamma distribution for the noise parameters is adopted so that the predicted and the updated intensities can be represented by mixtures o f Gaussian–Gamma terms. As the target state and the noise parameters are coupled in the likelihood functions, the variat ional Bayesian approach is applied to derive approximated distributions so that the updated intensit y is represented in the same form as the predicted intensity and the resulting variational Bayesian PHD(VB-PHD) algorithm is recursive.Chapter 4 is extended target tracking. And extended target tracking focuses on the method to easily partition the measurements into a number of subsets, each of subsets is supposed to contain measurements that all stem from the same source in principle, but in order to facilitate the research, we use the relatively simple Mahalanobis distance separation. The simulation experiments show that the proposed VB- PHD filter tracking effect is better than GM-PHD filter.