Comparison of fusion methods for multiple target tracking

Fusion of data from multiple sensors for the purposes of tracking multiple targets is considered. Two methods of fusion are considered. The first uses all measurements to perform tracking while the second adds a clustering step which attempts to remove clutter prior to tracking. Clustering is performed using a sequential Bayesian analysis and tracking is performed using a Monte Carlo approximation to the multiple hypothesis tracker (MHT). In performance analyses involving up to 25 targets, tracking with all measurements performs significantly better than tracking with the output of the clustering algorithm.