18th International Conference on Information Fusion (Fusion), Washington, Kiribati, 6 - 09 Temmuz 2015, ss.870-877
A fusion methodology for tracks represented by Gaussian mixtures is proposed for distributed maneuvering target tracking with unknown correlation information between the local agents. For this purpose, Chernoff fusion is applied to the Gaussian mixtures provided by the local interacting multiple-model (IMM) filters. Chernoff fusion of Gaussian mixtures is achieved using a recently proposed method in the literature involving a sigma-point approximation. The results show that the fusion of Gaussian mixtures in a distributed maneuvering target tracking scenario brings a moderate improvement over fusing only moment matched Gaussian densities.