Research On Extended Targets Tracking Based On PMBM Filter In Complex Environment

With the development of sensor technology and the perfection of shape modeling theory,the extended target tracking technology comes into being.Because it overcomes the defect that the point target tracking technology can't provide the target shape information,it has attracted much attention from scholars and become a research hotspot.In view of the advantages of random set filtering theory in filtering accuracy and modeling of targets,the effect of target tracking can be improved by integrating the extended target shape modeling theory with the random set filtering technology.Based on the extended target poisson multibernoulli mixture(PMBM)filter,this thesis studies the extended target tracking method in the complex environment.Specific research contents are as follows:Firstly,aiming at the problem that the existing PMBM extended target tracking algorithm has a low accuracy in estimating the extended shape of elliptic targets in the clutter environment,a PMBM extended target tracking algorithm based on the image moment elliptic random hypersurface is proposed.The algorithm combines the excellent mathematical characteristics of image moment and the advantage of target range of elliptic random hypersurface modeling,which makes the estimation of elliptic target more accurate.Furthermore,in order to estimate the extended shape of non-ellipsoid targets effectively,PMBM extended target tracking algorithm based on gaussian process regression is proposed.The algorithm adopts a separate estimation strategy and uses PMBM extended target tracking algorithm to estimate the centroid state of the target.Meanwhile,the extended shape of the target is inferred in real time by gaussian process regression technique.Simulation results prove the advantages of the two algorithms in estimating the extension.Secondly,aiming at the problem that it is difficult to accurately obtain the background model parameters such as detection probability and clutter in advance,a PMBM extended target tracking algorithm which can estimate detection probability and clutter is proposed.The algorithm first uses the finite mixed probability model to represent the spatial distribution of clutter,and the statistical characteristics of detection probability by beta distribution modeling.Then,the EM algorithm or gibbs sampling algorithm are used to solve the parameters of the mixed probability model.Finally,Finally,the obtained clutter parameters are introduced into the filter to realize the filtering recursion.Simulation results show that the proposed algorithm can effectively track the target in both uniform and non-uniform clutter scenarios.Finally,aiming at the problem that the performance of the extended target tracking algorithm is degraded due to the inaccuracy of the measurement noise covariance,the PMBM extended target tracking algorithm based on the inverse gamma distribution and the PMBM extended target tracking algorithm based on the inverse wishart distribution is proposed respectively.The measurement noise covariance is defined as a state variable,and add it to expand the list of target state form the united state variables,and then use variational bayes technology to constitute a joint probability density function of the effective approximation,the inverse gamma distribution and inverse wishart distribution can effectively characterize the statistical properties of the measurement noise.The simulation results verify the effectiveness and stability of the proposed two algorithms for tracking the multi-extension targets under the covariance of unknown measurement noise.