A Bayesian Approach to Estimating Detection Performance in a Multi-Sensor Environment

Abstract : This report presents the results of Operational Research support to maritime surveillance operations in the Canadian maritime approaches. The principal development is a Bayesian method to estimate the performance of sensors in a way that can be applied during ongoing operations. This novel method does not require knowledge of the sea-truth to evaluate sensors. A method to estimate sea-truth is also presented, which provides a new and unique analysis capability for the estimation of surveillance gaps. The series of new metrics and methods are implemented in a variety of operational tools for analysis of the recognized maritime picture (RMP), which support new operational procedures and processes. The desired result is a more robust and optimized maritime surveillance capability.