Final Report,

Activity under this contract has centered on two general problem areas related to the airborne detection of underwater targets. The first utilizes anomalies in the earths magnetic field to detect the presence of a target, while the second is a detection problem associated with returns from an optical radar. Two models and approaches for the detection of targets by means of measuring magnetic anomilies have been studied. In the first, the nonlinear equations relating position to its magnetic signal are used directly. After a preliminary study of some recursive estimation techniques, a more detailed study of using an extended Kalman filter as an off-line sequential estimator was undertaken. A novel idea in this approach was to process the data with random samples rather than in the causal sequential order in which they occur. As discussed below in Section A.1, this enhances convergence of the parameter estimates since it avoids directly processing correlated data. The second model is a linear version of the nonlinear equations, and a combined detection - estimation problem is formulated. As discussed in section A.2, approximate non-Gaussian and robust state estimation combined with sequential likelihood ratio processing appears to be an attractive approach for detecting targets in a switching environment, and also where models are not very refined.