Research On Algorithms Of Multiple Extended Targets Tracking Based On Random Hypersurface Model

With the improvement of sensor resolution,one more resolution units are occupyied by sigle target,it is more reasonable to regard this target as an extended target rather than a pointed target.In each sampling period,an extended target raises measurements that stem from dierent locations,named measurement sources.Most traditional multi-target tracking formulations are no longer suitable because of intractable complexity and discard of the advantages of high resolution sensors.In recent years,some random finite set(RFS)filters have been developed,and draw lots of attentions because of its directly tracking the multi targets with low computational complexity.In order to improve the robustness and precision of the PHD filter and the CBMe MBer filter,it is desired to estimate the target extent in addition to the kinematical state,using random matrix model and random hypersurface model.The main research contents of this thesis are as follows:1.Combined with the CBMe MBer filter,an algorithm of extended object tracking using random matrix is proposed for ellipse extended object,named the GGIW-CBMe MBer filter.It models the measurement rate as a Gamma distribution,the kinematical state as a Gaussian distribution while the extension state as an inverse Wishart distribution.With target predicted and updated in this way,a simultaneously estimate is obtained.The simulation shows that the proposed algorithm can effectively estimate the measurement rate,the kinematic and the extension state,and it is more accurate than the GGIW-PHD filter.2.The measurement sources modeled by the elliptic random hypersurface model can be easily embedded into the CBMe MBer filter,and the RHM-GGM-CBMe MBer filter is proposed.In this algorithm,the possible set of measurement sources can be obtained by a scaling factor,and then the measurement source is selected randomly.It is an appropriate expression of lacking of the prior information.The choice of scaling factor does not depend on the extended shape of the target,thus this is not hierarchical.The shape parameters of the ellipse are incorporated into the kinematical state vector,so that the processing of the matrix is avoided.Simulations for tracking of an unknown number of targets in the presence of clutter are attached,which show the RHM-GGM-CBMe MBer filter outperforms the RHM-GGM-PHD filter and the GGIW-CBMe MBer filter.