Research On Methods Of Multiple Extended Targets Tracking Based On Random Finite Set

Multiple Target Tracking(MTT)has always been a research difficulty in the field of target tracking.Its purpose is to estimate the state and the number of targets,given the uncertainty of both the targets and the measurements.With the continuous development of electronic technology and the improvement of sensor accuracy,one target occupying multiple resolution units will produce multiple measurements per time step,this target is treated as extended target.Since extended target will produce more than one measurements per time step,it is great difficult to determine the corresponding relationship between measurements and targets by utilizing traditional data association methods.In recent years,the multiple targets tracking method based on random finite sets has been widely concerned,because it avoids the complicated data association problem.Therefore,the research on the multiple extended targets tracking method in this thesis is based on the theory of random finite set.The main contents of this thesis are as follows:1.A novel measurement partition algorithm of Fuzzy c-means(FCM)based on adaptive density is studied to address the partition measurement set of multiple extended targets.Firstly,The algorithm reduces the computational complexity and improves the accuracy of measurement partition by adaptive gate,and only the measurement sets is preprocessed to remove clutters from the measurements.Secondly,initial cluster center is generated by the density function,and the detailed process of FCM based on adaptive density is elaborated.Finally,compared with the existed partition algorithm like Kmeans++ partition,the simulation results show that the proposed algorithm obtains more accurates results,as well as lower time cost.2.For the particle degradation problem in the implenmemtation of extended target particle probability hypothesis density(PHD),this thesis studies multiple extended targets cubature kalman particle PHD.The algorithm accounts adequately the latest measuring information in the prior updating phase,and utilizes cubature kalman filter to develop the importance density function of multiple extended targets particle PHD.As a result,the particles can be sampled in the state space corresponding to the actual measurement of the extended target.So that the sampled particles move to the peak region of the likelihood function,which can more accurately approximate the posterior intensity of the multiple extended targets andimprove the tracking performance of the multiple extended targets particle PHD.The simulation results show that the presented algorithm mitigates the particle degradation and improves the overall performance.3.This thesis studies the multiple extended targets tracking problem based on random matrix under nonlinear cases.Firstly,this thesis studies the modeling and tracking method based on the random matrix extended target.The method is used to model the shape of the extended object by the ellipse presented the symmetric positive definite random matrix.Secondly,using the modified unbiased converted measurement(MUCM)method,the radar measurement is expressed by the coordinate transformation as the pseudo-linear form in the Cartesian coordinate system,and obtaining the variance of conversion measurement error by statistical method.Then,the concrete realization process of Gaussian inverse Wishart Probability Hypothesis Density(GIW-PHD)filter algorithm is described in detail based on MUCM.And obtains the motion state and shape estimation of multiple extended targets by recursiving GIW-PHD algorithm.Finally,The simulation results show that the effectiveness of the algorithm.