New and better algorithm for multitarget tracking in dense clutter

JPDA (joint probability data association) algorithm and variations of JPDA - JIPDA and IJPDA - all appear to suffer from using a feasible rule that is not quite in accord with real environment. We propose a new algorithm, generalized probability data association algorithm (GPDA), which is based on a new feasible rule, and we hope it is better for multitarget tracking in dense clutter. The paper offers a new feasible rule, which considers that measurements and targets may be used repeatedly. The paper takes the careful definition of a generalized joint event as being composed of two dummy events. This paper discusses how to use Bayes' rule to calculate the marginal probability β_it. Computer simulation results show that the algorithm we propose for multitarget tracking in dense clutter is better in three aspects: lower tracking ratio loss, higher tracking precision and less computational burden.