Distributed-system detection in correlated clutter

Abstract

Distributed sonar systems decide if a target is present through use of a rule that combines information from each individual sensor. The most basic of these are the ‘‘and’’ and ‘‘or’’ data fusion rules. Analysis of these rules has typically been restricted to cases where the target is detected on some number of sensors, and the false alarms are assumed to arise from processes that are statistically independent from sensor to sensor. In active sonar systems operating in clutter‐dominated areas, this assumption may be far from the truth. Active sonar clutter often has its origins in scattering from physical objects (e.g., shipwrecks, fish schools, rock outcroppings, or mud volcanoes). Particularly when the scatterers are small in terms of wavelengths, this leads to false alarms that can be highly correlated from sensor to sensor. In this work, clutter scattering is assumed to arise from a finite number of scatterers with an exponentially‐distributed size, which would lead to a K‐distributed matched‐filter envelope. The performance of the ‘‘and’’ and ‘‘or’’ fusion rules is then evaluated for two sensors with partially overlapping resolution cells. The resulting detection statistics from the two sensors are not independent, necessitating approximation of the joint distribution function.