Optimal fusion performance modeling in sensor networks

This paper studies sensor network surveillance performance at the automatic tracker output. In particular, we develop a simple analytical model for tracker performance, where the interest is in a compact representation of the impact of sensor revisit time. This model, combined with a previously developed contact fusion model, allows for an analysis of two fusion architectures: a standard centralized tracking configuration and a ldquofuse-before-trackrdquo two-stage approach. We find that the latter architecture is promising, though in the specific instantiation considered here there are performance limitations when targets are in close proximity. Additionally, in the context of centralized tracking, we study optimal sensor thresholding and sensor network size for a given false track rate.