Efficient Algorithm for Processing SAR Data Based on GPU

Abstract

Data processing is time-consuming in the field of Synthetic Aperture Radar (SAR). Graphics Processing Units (GPUs) have tremendous float-point computational ability and a very high memory bandwidth, and the developing Compute Unified Device Architecture (CUDA) technology has enabled the application of GPUs to general-purpose parallel computing. A new method for processing SAR data using GPUs is presented in this paper. Compared with the nominal GPU-based SAR processing method, the number of data transfers between the CPUs and a GPU is reduced from 4 to 1, and the CPUs are exploited to cooperate with the GPU synchronously. By using the proposed method, we can speed up the data processing by 2.3 times, which is verified by the testing with simulated SAR data.