Tracking of group targets using extended kalman filter and interacting multi-model

Abstract In order to solve the challenge of joint estimation of state and shape in multi-target tracking in complex environments, this paper proposes an innovative method based on an interacting multi-model extended Kalman filter. By constructing a joint state space containing target motion state and elliptical shape parameters, a pseudo-measurement update mechanism based on second-order/fourth-order center distance is designed to achieve accurate estimation of nonlinear shape parameters. Combined with the interacting multi-mode algorithm framework, the problem of target motion mode switching is effectively handled. Simulation experiments show that in the accelerated motion scenario of group targets, the proposed method reduces the Gaussian Wasserstein distance error compared with the traditional random matrix method and standard extended Kalman filter, and improves the tracking accuracy in the target direction mutation stage.