GPU-based parallelized time-domain back-projection processing for Agile SAR platforms

Agile SAR platforms such as an automobile require a flexible SAR processing scheme to account for nonlinear sensor trajectories during the synthetic aperture. In this contribution, a parallelized implementation of a time-domain back-projection SAR focusing algorithm based on NVIDIA's CUDA GPU computing framework is presented and discussed using a car-borne SAR data set. The processing performance is assessed using different hardware. In addition, a pre-processing scheme is described that allows for full 3-D motion compensation, yet staying conveniently in conventional slant-range/azimuth geometry of single-look complex SAR images.