Performance and Computational Trade Analysis for Low-SWaP Synthetic Aperture Radar Application

摘要

Airborne radar platforms with low size, weight, and power (SWaP) minimize cost and operational risk but also place constraints on the complexity of the algorithms used to process the received data. Producing high quality radar images requires sophisticated algorithms that provide accurate and useful information regarding the scene of interest. Imaging algorithms that provide the highest quality synthetic aperture radar (SAR) imagery come with a steep computational price: O(N~3) operations are required to produce an N pixel × N pixel image; imaging algorithms that require fewer computations introduce errors of varying degree that result from approximations to the signal model. Several sub-optimal SAR imaging algorithms were evaluated to quantify the impact of these approximations for various system parameters and scene sizes. A comparison of image formation times on two low-power graphical processing units (GPUs) provides a realistic comparison of algorithm complexity. Additional algorithms and techniques in the end-to-end image formation process that provide favorable trades between performance and computations are identified, and design challenges specific to a SWaP-limited platform are addressed.