Synthetic Aperture Radar Processing on GPU

Abstract

Synthetic Aperture Radar (SAR) system operates in all weathers conditions to produce imagery of a terrain through reflections from that scenario during a manure. For verification of applied algorithms, it is a tedious task to operate SAR in real world environment and gather data. SAR data for a typical scenario can be synthetically generated but it takes lot of valuable time up to hours. As SAR processing is a two dimensional complex problem, debugging and verification of post processing on such a huge data is a tedious task too. In this thesis, whole simulation of SAR imagery i.e. data generation and post processing, is optimized using GPU by parallelizing data synthesis and algorithm applied. Modifications are made in applied algorithm for motion error correction, and SAR autofocus. Results show promising improvements in synthetic data generation catalyzing reduction in time from hours to a few minutes. Optimizations are enhanced to such extent that real time SAR imagery is achieved.