Research Of Multi-target Tracking Algorithm Based On Random Finite Set

Multiple target tracking(MTT) technology is one of the most important technology in the field of multi-source information fusion, which has been widely used in civil and military fields. As the uncertain factors such as the number of tracking targets in the process over time, which can increase the complexity and decrease the tracking accuracy for traditional multiple target tracking algorithm. In order to solve this problem, random finite set (RFS) multi-target tracking method is adopted in the paper. The method is applied to the multiple target tracking algorithm on linear and nonlinear conditions respectively. The main content is as follows:(1)Aiming at solving the problem that the solution of PHDF is difficult to get closed due to recursive equation of multiple integral operation. The characteristics of gaussian mixture model(GMM) is adopted in the condition of linear gaussian. The gaussian mixture probability hypothesis density (GMPHD) filtering algorithm is proposed by using gaussian mixture composition instead of probability to posterior probability hypothesis density (PHD). Finally, the simulation experiment is completed in MATLAB. The results showed that the GMPHD filter algorithm is not only suitable for the target tracking that the number changes over time and also can solve the problem of tracking trajectory cross in the process of multiple target tracking.(2) Aiming at solving the problem that GMPHD is not suitable for the nonlinear systems, EK-GMPHD is presented with the theory of local linearization. The Linearization method is applied to the nonlinear function through Taylor series expansion and get the approximation. The Jacobian matrix of the state transfer function and the likelihood function are calculated through partial derivative operation process. In order to recursive estimate the number and the status of targets,the Jacobi matrix takes the place of the state transition matrix and measure matrix in GMPHD filter algorithm. The simulation results show that EK-GMPHD algorithm can solve the MTT problem with weak nonlinear condition.(3) The EK-GMPHD filter algorithm can generate large linear deviations in the high degree of nonlinear condition. The gaussian mixture probability hypothesis density particles (P-GMPHD) filtering algorithm is adopted to solve the problem. First, the Monte Carlo method (Monte Carlo) is applied to the integral item PHD filter algorithm to get closed solution approximate calculation and the tracing prediction equations are derived. The simulation results illustrate that the P-GMPHD filter algorithm is suitable for the high degree of nonlinear MTT system.Through the research content of this paper, we can conclude that the finite set statistics (FISST) theory applied to the MTT algorithm can not only reduce the amount of calculation but also improve the tracking accuracy.