Research On Multiple Maneuvering Extended Target Tracking Algorithm Based On PHD

Target tracking technology is a method to estimate the state information of the target by using the inaccurate measurement information with clutter.Target tracking has been one of the hot spots in the field of data fusion and it is widely used in the military and civil fields.In the traditional target tracking algorithms,it is assumed that the target corresponds to only one measurement at each time,but this assumption is no longer applicable in the case of high resolution sensors.In the actual application scenes,a single target may correspond to multiple sensor measurements at each time interval,such target is called extended target.In recent years,the extended target tracking technology has been widely concerned.Under the framework of PHD filter,this paper proposes some related improvements on the ET-GMPHD filter algorithm for improving the tracking performance of the extended maneuvering target.Furthermore,the related research work is carried out for the problem that the improved algorithm can not obtain the complete trajectories of each extended target.1.In this paper,the extended target tracking algorithm realized on Gaussian mixture and the common maneuvering target tracking methods are studied.The modified input estimation method is introduced to ET-GMPHD algorithm to improve the tracking accuracy of multiple extended maneuvering target.The proposed ET-MIE-GMPHD algorithm derives a new filter gain by constructing a cross-covariance matrix based on extended process noise and extended measurement noise.When the extended target is maneuvering,the new filter gain is corrected in real time according to the cross-covariance matrix,in this way,the multiple extended maneuvering target can be tracked.Simulation results show that the algorithm has better tracking performance in the tracking of weak maneuvering targets,which indicates that the algorithm has certain practical application value.2.A multiple extended maneuvering target tracking algorithm based on Interacting multiple model is studied.Although the ET-GMPHD algorithm proposed by Granstr?m under the liner Gaussian condition can effectively track the multiple extended targets,but the performance will decrease for the multiple strong maneuvering targets tracking in the clutter environment.In this paper,an interacting multiple extended Gaussian mixture probability hypothesis density filter algorithm is proposed based on the ET-GMPHD filter with the idea of IMM.In contrast to the ET-MIE-GMPHD algorithm described above,the proposed algorithm can track the multiple extended targets in the strong maneuvering scenes.3.Although the proposed two multiple extended maneuvering target tracking algorithms can realize the multiple extended target tracking under the scenes of weak maneuver and strong maneuver,but the trajectories of each target can not be obtained.A track management method based on Gaussian component labeling method for multiple extended maneuvering target is proposed.Simulation results show that this method is feasible for the management of the extended maneuvering target trajectories.