Tracking of group targets using multiple models GGIW-PHD algorithm based on best-fitting Gaussian approximation and strong tracking filter

Gamma Gaussian inverse Wishart probability hypothesis density (GGIW-PHD) filter algorithm is always used for tracking group targets with unknown number and variable measurement rates in the presence of cluttered measurements and missing detections. Aiming at the defect that the tracking error of GGIW-PHD filter algorithm will increase greatly in the maneuvering stage, a multiple model GGIW-PHD (MM-GGIW-PHD) algorithm is proposed in this paper based on the best-fitting Gaussian approximation and strong tracking filter. Firstly, on the basis of measurement set partition, the best-fitting Gaussian approximation method is proposed to implement the fusion of multiple models in the PHD predict stage. And a fading factor of strong tracking filter is proposed to correct the predicted covariance matrix of the GGIW component. Then, the estimation of kinematic state and extension state are deduced in the frame of multiple models. The probability of different tracking models is updated by the modified likelihood functions. The simulation results show that the MM-GGIW-PHD algorithm based on best-fitting Gaussian approximation and strong tracking filter can decrease the tracking error of group targets in the maneuvering stage and treated with the combination/spawning of group effectively.