A multi-array bearing-only fusion framework for passive underwater multi-target localization

In strong interference underwater environments, multi-array multi-target bearing-only localization faces two major challenges: ambiguous measurement association and ghost points from incorrect bearing-line intersections. This paper proposes an effective fusion framework to address both issues. For the former, other than geometric distance, such as nearest neighbor, we propose to use Fisher information with statistical-theoretical guarantees as an optimal measurement association (OMA) metric. For the latter, we propose a multi-target maximum likelihood (MML) estimator by constructing the objective function with probability hypothesis density. Both simulation and sea trial experiments show that our MML-OMA framework suppresses ghost points and improves localization accuracy. Since tracking is usually required after localization in the detection processing chain to estimate the target kinematic state, we further introduce multiple hypothesis tracking as a downstream process in the sea trial experiments to evaluate localization performance. Compared with the traditional localization method, the proposed method significantly improves downstream tracking performance.