An Information-Theoretic Approach to Complementary Information Fusion
James Klupacs, Amirali Khodadadian Gostar, Alireza Bab‐Hadiashar, Reza Hoseinnezhad · 2023
This paper presents the challenges of divergence-based sensor fusion methods in multi-target tracking applications where sensors have different fields of view. We propose an alternative approach to GCI fusion that is particularly devised for the fusion of labeled multi-Bernoulli densities. In our approach, we present the observation that fusion could be separately performed for each object label over the single-Bernoulli densities associated with that label, and the fusion weights could then be tuned according to the detectability of the object by each sensor. The resulting solution is tested in a challenging numerical experiment where up to 10 vehicles are moving and tracked by various dynamic and static sensors. The results show the advantages of our proposed method when compared to GCI fusion and the complementary fusion methods.