Database Expansion Issues

State-of-the-art radar combat ID (CID) must address target identification in complex environments, often with very limited measured training data. Consequently, the algorithms frequently rely on data computed via numerical models, as a function of required sensor conditions (bandwidth, target-sensor orientation, polarization, etc.). Such data are usually computed using approximate high-frequency numerical models, because targets of interest are often very large relative to wavelength. As a consequence of the number of targets and poses that must be considered, even these approximate numerical models require substantial computational resources. It is therefore essential that one develop techniques to prioritize where computational efforts should be directed, for improving the accuracy of numerically-generated data and overall CID system performance. In addition, by identifying the key phenomenological primitives (features) driving CID performance, computations can be accelerated significantly by developing simpler models linked to the key scattering features. The Phase II goals are to extend the Phase I methodologies associated with identifying key features and phenomenology that drive CID performance to targets and databases of interest to AFRL, and the principal Phase II products are the tools required to assist in these efforts. Commercial applications include medical imaging, land-use assessment, and DoD modeling and analysis software.