FPGA Implementation of Two SAR Autofocus Algorithms

Abstract

SAR (Synthetic Aperture Radar) imaging processing is of high computations and its imaging algorithms are often complicated. How to implement SAR imaging efficiently and in real time is a deserving research. Autofocusing is an essential chain of SAR processing and is also computation-consuming. We know that FPGA (Field Programmable Gate Array) has many advantages such as being of large scale, high-performance and reconfigurable and so on. So we would like to utilize FPGA to implement SAR autofocus algorithms in real time. In this paper, implementations of two popular autofocus algorithms of SAR imaging, namely, PGA (Phase Gradient Autofocus) and PAST (Projection Approximation Subspace Tracking) are discussed in detail based on KintexTM-7 FPGA of Xilinx Corporation. State machines and hardware mappings of each are designed respectively with pipelined processing and floating IP core which ensures all the necessary computations of floating numbers efficiently. Our experimental results show that it takes both algorithms about 0.5 seconds to completely process a defocused SAR image of 2k*2k 64-bit samples with system clock 200MHz, which authenticates the valid and credible real-time implementations. This work lays the foundation of the FPGA combination of autofocus algorithms with basic imaging algorithms to construct a complete real-time SAR processing system in the future.