Performance evaluation of multiple-target tracking systems

To evaluate the performance of real multiple-target tracking systems, instead of their simulative systems, the signal origin of each track should be determined in the different form because the data association in real systems is absolutely unknown. In addition, the calculation of performance indices should be suitable for the case of variable sampling interval. Here the state estimate of each track is projected into measurement space and the statistical distance between this projection and each simultaneous measurement is calculated. Hence the signal origin of that track can be determined via the rule of Nearest Neighbor. Based on the relationship matrix with respect to track and signal origin, a new method to calculate a suit of performance indices for both TAS and TWS modes is proposed. This method has been adopted in our TrackEnviro - a evaluation system for the multiple-target tracking performance of real phased array radar.