Extension of Nonlinear δ-generalized labeled multi-Bernoulli Filter in Multi-Target Tracking

This paper proposes two nonlinear delta-generalized labeled the multi-Bernoulli (delta-GLMB) filters by combining quadrature Kalman (QK) and cubature Kalman (CK) with delta-GLMB litter for nonlinear Gaussian mixture (GM) multi-target model. And the QK-GM-delta-GIMB and CK-GM-delta-GLMB filters are compared with other two existing nonlinear delta-GLMB tracking algorithms, which are respectively based on the Extended Kalman Filter (EKF) as well as the Unscented Kalman Filter (UKF), about tracking performance for multi-target tracking (MTT) in the case of different clutter densities as well as detection probabilities. Moreover, the GM implementation process of the four filters is compared with the SMC simulation process of the delta-GLMB filter in the nonlinear Gaussian model. Simulation consequences demonstrate that both the QK-GM-delta-GLMB and CK-GM-delta-GLMB filters are effective nonlinear tracking methods, and we can conclude that the QK-GM-delta-GLMB filter is a competitive approach that achieves better tracking performance at an acceptable time cost for mildly nonlinear multi-target motion scenarios.