A Method for Track Association Using Dynamic Programming

Track association is one of the key problems in modern distributed multi-sensor information fusion systems design, which paves the way for the subsequent combination of state estimates. A distributed data processing structure is preferred in such a distributed system, which results in a lot of advantages over centralized architecture. Track association generally contains two steps: establishing an association metric to evaluate each track-to-track association hypothesis, and the selection of the best assignment between sets of tracks based on association metric. Feature assisted tracking often can't be realized in lots of situations, especially in tracking to opponent target, in that the features can't be acquired effectively. A track association method is presented based on exploiting the inherent nature of the targets' tracks using only kinematic data, which was implemented by dynamic programming. The accumulated sequential statistics' taking account of the cross-covariance were adopted in formulating the association metric, providing robust similarity description to track-to-track association hypothesis in complex target environment. Then the best assignment between sets of tracks was achieved by solving an optimization problem using dynamic programming. Numerical simulations are conducted to show the performance of the presented method.