A real-time multitarget tracker by adaptive hypothesis testing for airborne surveillance systems.

Abstract

A multitarget, multisensor tracking algorithm is formulated specifically for the environment of an airborn surveillance system. The algorithm divides report/track pairs into unique associations and clusters of cross associations. The clusters are solved independently via an adaptive hypothesis testing framework (multiway tree) using a maximum likelihood test. An intelligent scheme is provided to contend with hypothesis tree overflow. Tracks are allowed to split in response to ambiguous associations and true trajectories determined probabilistically outside the hypothesis tree. Performance comparisons are made with nearest neighbor and split track algorithms.