Performance of cued target acquisition systems: Impact of automatic target recognition

This paper addresses the influence of automatic target recognition (ATR) subsystem performance on cued target acquisition system performance. Performance parameters used in surveillance systems, such as probability of detection and probability of correct classification/declaration, are incomplete in representing the performance of cued acquisition systems, which commit resources to an object designated by an external cue rather than simply issuing surveillance reports. Target acquisition is posed as a task of selecting from among multiple detected objects based on a joint likelihood ratio test indicating the consistency of their sensed position, velocity, and signature with an external cue. The acquisition function thus computes an order statistic, over all competing objects, of this joint likelihood ratio. The resulting expression for probability of correct acquisition is then used to analyze performance gains achieved by fusing ATR into a kinematics-only acquisition function. Monte Carlo simulation results indicate the accuracy of simplifying approximations and the system performance uncertainty introduced when ATR performance is specified only by classification error rate.