Multi-Hypothesis Sonar Tracking

This paper introduces a multi-hypothesis multistatic sonar tracker for undersea surveillance. Multistatic sonar increases the data rate and has the potential to improve surveillance capabilities, provided effective target tracking is performed. Our multihypothesis tracker includes features not generally found in other multi-hypothesis trackers. Data association is based on an efficient linear programming approach, to which we introduce a novel modification that improves track continuation. We use equality constraints in the LP, and tracks are removed when they fail a confirmation criterion. Short duration tracks are classified as false and removed. System and measurement uncertainties are reflected through multistatic contact covariances. This uncertainty impacts the data association hypotheses that are considered, as well as their log-likelihood scores. We test the improved performance of this tracker over our earlier baseline tracker, with a number of benchmark examples of interest and through Monte Carlo evaluation.