Multisensor and Multitarget Tracking Based on Generalized Covariance Intersection Rule

Abstract

Distributed multitarget tracking (MTT) is suitable for sensors with limited field of view (FoV). Generalized covariance intersection (GCI) fusion is used to solve the MTT problem based on label probability hypothesis density (PHD) filtering in this paper. Because the traditional GCI fusion only has good fusion performance for the targets in the intersection of each sensor’s FoV, and the targets outside the intersection range would be lost, this paper redivides the Gaussian components according to the FoV and distinguishes the Gaussian components of the targets inside and outside the intersection. GCI fusion is sensitive to label inconsistency between different sensors. For label fusion in the intersection region, the best match of labels is found by minimizing label inconsistency index, and then GCI fusion is performed. Finally, the feasibility and effectiveness of the proposed fusion method are verified by simulation, and its robustness is proved. The proposed method is obviously superior to local sensor and traditional GCI algorithm.