Clutter Estimation And Multiple Extended Target Tracking Based On Random Finite Set

Extended target tracking problem has always been a hot issue at home and abroad.Extended target tracking has the following characteristics:First,the target generates multiple measurement points in one sampling period,and these measurements are usually located in the same sensor(such as radar).Within the resolution unit,the difficulty of estimating the number of extended targets is increased;secondly,the shape of the extended target is usually unknown,and may even change with time,resulting in an increase in difficulty in estimating the shape of the target.Based on the theory of random finite sets,this paper firstly proposes the RFS-based clutter intensity estimation algorithm,then establishes the scattering point model of the extended target.Finally,a multi-extended target tracking algorithm based on labelled random finite sets framework under clutter condition is proposed.The specific research contents are as follows:(1)RFS-based unknown clutter intensity estimation algorithm.In many cases,target tracking is subjected to a dense,non-uniform,and time-varying clutter background.This will seriously deteriorate the tracking performance under an unknown clutter environment.Considering the detection rate and clutter rate of unknown clutter,a finite mixture distribution(FMD)is introduced to fit the unknown clutter distribution,and then the Gibbs sampling and Bayesian information criterion(BIC)are used to evaluate and estimate the clutter parameters.(2)Extended target scattering point modeling.With the continuous development of modern sensor technology,the increasing resolution of radar enables us to obtain multiple scattering point information from a single target,that is,an extended target generates more than one measuring points in one sampling period,which requires the establishment of an extended target scattering point model.Firstly,the extended target scattering point state is modeled by multi-Bernoulli RFS.The state is represented by a multi-Bernoulli distribution model with parameter set {(r(i),p(i)}i=1 M.It is assumed that the number of extended target scattering points satisfies the Poisson distribution,and the scattering point parameters are with Gaussian mixture model.Realization of extended target scattering point structure modeling by graph theory method.(3)A tracking learning algorithm for multi-extended targets under clutter condition.The algorithm mainly includes two parts:dynamic modeling of multiple extended targets and tracking estimation of multiple extended targets.By expanding the research content(1),this paper combines the GLMB filter to establish the finite mixture model of the extended target The Gibbs sampling and BIC criteria are used to derive the parameters of the finite mixture model to track the multi-extended target.The equivalent measurement of the extended target is obtained to replace the extended target measurements,and the extended target shape is modeled by ellipse approximation to realize the estimation of the extended target state and shape.