Frequency domain feature extraction from synthetic aperture radar data

Most feature recognition and classification algorithms for synthetic aperture radar (SAR) are done in the image domain, and thus do not explicitly exploit the scattering physics underlying the data. Feature extraction in the phase domain is based on electromagnetic scattering models which describe the sensor response to scene features. This idea of processing the SAR data using a filter matched to the phase history signature of a feature was introduced by Franceschetti (1997). Here, we extend the original concept for processing the raw signal to processing the spotlight mode SAR signal. We use a sensor model based on an experimental airborne SAR developed by MIT Lincoln Laboratory, the Lincoln Multimission ISR Testbed (LiMIT). To demonstrate the concept, line feature filters are applied to a simulation of a building from the LiMIT dataset.