Combinatorial problems in multitarget tracking¿a comprehensive solution

The paper presents a new solution to the combinatorial problems that arise when multiple targets are to be tracked simultaneously. In multitarget situations, the correspondence between targets and observations may be lost because of crossing trajectories, close-by targets, false alarms, missed detections etc. The number of data association hypotheses that are feasible and for which probabilities have to be calculated is extremely large in any but the smallest of such problems. We present a new algorithm for locating the N globally best hypotheses using a simple search method instead of extensive enumeration or integer programming methods.