Robust Target Identification Using a Generalized Likelihood Ratio Test

Abstract

The concept of deciding among a set of alternatives (or hypotheses) based upon the observation of a set of random variables has been a topic studied by statisticians for many years. This concept, known as hypothesis testing, provides a mathematically solid foundation to perform target identification. Target identification with known signatures can be easily formulated using Bayes hypothesis testing. However, a significant challenge lies in the need to accurately discriminate among known targets with only partial knowledge of target signatures. The lack of complete target signature knowledge results from the unknown orientation of the target and the dependency of the target signature on the target’s orientation. For practical purposes, it is important to derive efficient and reliable schemes to accurately identify the target without a priori knowledge concerning the target’s orientation.