Multi-Dimensional Automatic Target Recognition

This research will investigate the use for Automatic Target Recognition of data other than relative amplitude images obtained from Synthetic Aperture Radar (SAR) sensors. Unlike human image analysts who can quickly process only (relative) amplitude images, ATR systems have the potential to use physical models and other forms of target reference data to process additional types of collected and processed data. Additional data types include absolute amplitude images, polarimetric images, multi-kernel images, symbolic feature segmentations, and prior models of the scene including multilook data. The desirability of using such data types has been widely acknowledged, but techniques have not advanced enough to be suitable for incorporation into development programs headed for near-term deployment. The proposed effort will develop a flexible Multi-Dimensional Automatic Target Recognition (MDATR) system for Bayesian modeling and associated ATR, populate the system's model and algorithm databases to work with multiple types of SAR-derived data, and test the resulting system performance.