Multi-target joint tracking and classification based on joint decision and estimation

For the multi-target joint tracking and classification (JTC) problem, the interdependence between tracking and classification must be considered. This study introduces a conditional joint decision and estimation (CJDE) frame into a probability hypothesis density (PHD) filter to resolve multi-target JTC problems with varying target numbers and observation uncertainties. Specifically, the optimal subpattern assignment (OSPA)-based estimation and decision intermediate-cost of multi-target is defined. Moreover, conditional PHD update equations are developed for different target class decisions to calculate the estimation and intermediate costs, and the optimal multi-target CJDE result is derived based on the minimum Bayesian risk criterion. Illustrated examples of the proposed CJDE-PHD method are presented, numerical simulations are compared with the PHD filter and two-stage algorithm, and the effectiveness of the proposed method is demonstrated.