FOPEN Radar ATR Using Superresolution and Eigentemplates

The detection and recognition of man-made objects under foliage conditions requires the use of VHF/UHF ultra-wideband (UWB) radars such as P-3/SAR system developed by ERIM for NAWC/AD and DARPA. At low frequencies and long wavelengths, most of the scatters are in the Rayleigh scattering region and their RCS is low, which enables foliage penetration. However, this also results in a small number of large scatterers on target of interest. Automatic Target Recognition and Detection (ATR/D) using FOPEN SAR needs to utilize higher dimensional features such as polarization, absolute intensity, angle-diversity, frequency-diversity and other multi-kernel attributes of SAR imagery. In addition, since the resolution at low frequencies is degraded, super resolution techniques hold great promise for enhancing ATR performance. There is also a great need for dimensionality reduction in the design of ATR algorithms since the use of higher dimensional attributes results in multidimensional target templates. In this proposal, we describe an efficient ATR approach based on the use of eigentemplates which have been found to be very successful in face recognition and have been shown recently to be close to full template matching in SAR ATR performance. The specific Phase I technical objectives are: 1. aquisition of UWB FOPEN SAR simulated and real data. 2. Resolution enhancement using multidimensional Super resolution algorithms. 3. Development of Eigentemplates and associated ATR algorithms using higher dimensional target attributes. 4. Testing and evaluation of Supper resolution and ATR algorithms on simulated and real data. University of Maryland (Prof. Chellappa) will provide technical support for the SBIR effort.