Observation bootstrapping CDK-GMPHD filter based on consensus fusion strategy

Abstract Aiming at the adverse effect caused by the nonlinear, the randomness of observation noise and the uncertainty of external disturbances on the precision of state estimation and target number identification in multi-target tracking system, we propose a novel observation bootstrapping CDK-GMPHD filter based on consensus fusion strategy in this paper. Firstly, for the nonlinear of estimated system, we introduce central difference Kalman filter(CDKF) into the realization framework of Gaussian mixture probability hypothesis density (GMPHD) filter, and designs a new improved GMPHD filter based on CDKF. Secondly, considering the extraction and utilization of redundancy and complementary information form multiple observations can improve the stability and reliability of observations, we construct the bootstrapped observation set by the data assimilation technique. And on this basis, combined with the realization mechanism of membership function in fuzzy set theory, the consensus distance and the consensus matrix are further designed on the basis of confidence distance. Finally, according to the weighted fusion strategy, we realize the effective utilization of bootstrapped observations. The theoretical analysis and experimental results show the feasibility and efficiency of proposed algorithm.