GPU Acceleration of Synthetic Aperture Radar Processing
Abstract
We will investigate the use of GPUs to increase the efficiency of synthetic aperture radar (SAR) processing algorithms with complex survey geometries. Traditionally, when processing efficiency is a priority in a SAR algorithm, frequency-domain techniques that employ FFTs are used to accelerate performance but rely on approximations of the survey geometry. More specifically, these technique are limited to simple linear geometries that do not completely represent the actual survey configurations in which the radar data were collected resulting in degraded signal-to-noise ratios and resolution. More complex time-domain algorithms do not require such approximations but are considerably less efficient. By integrating the use of GPU with time-domain SAR algorithms, we can retain a better representation of the survey geometry while still achieving improved efficiencies through hardware parallelization. We will apply these techniques to airborne radar sounding data collected over the Greenland and Antarctic ice sheets.
Intellectual Merit
This project will develop efficient SAR algorithms that can be used to focus radar data collected over complex geometries.
Broader Impact
The data produced as a result of this project will provide improved knowledge of the bed conditions of the Greenland and Antarctic ice sheets. These are key inputs to ice sheet models that are used to predict the response of the ice sheets and their contribution to sea level rise in a warming climate.
Use of FutureGrid
Future grid will provide access to GPU processing environment.
Scale Of Use
Couple days per week.