Research On SAR Autofocus Algorithms And FPGA-based Implementations

Autofocusing is an essential step in the SAR (Synthetic Aperture Radar) imaging which studiesboth RCM (residual range cell migration) and azimuth phase errors after motion compensation, whichis a supplement to the onboard motion unit. This paper will investigate the autofocus algorithms andFPGA-based hardware implementation.Chapter1first gives a brief introduction to the research background and its meaning. Then itsummarizes several popular autofocus algorithms, including their advantages and disadvantages andpresents the current state of FPGA (Field Programmable Gate Array) based SAR development. Finallyit shows the whole contents of each chapter.In chapter2, several autofocus algorithms aimed to estimate the1D (One-Dimensional) azimuthphase errors are analyzed in detail. At the beginning, the processing step and estimate performance ofPGA (phase gradient autofocus) which is the basis of other algorithms are discussed. Later, weintroduce an updated algorithm called QPGA (Quality PGA) which is to accelerate the convergencespeed of phase error estimate of PGA. Based on PGA, we cope with the defocus problem of long CPI(coherent processing interval) and discuss three algorithms namely, PCA (phase curvature autofocus),PGA-LS and PGA-MD, respectively from the aspects of principal deduction, performance analysisand real data processing. The results show that the full-aperture PGA and PCA algorithms fail to copewith the long-CPI defocus problem. While PGA-LS and PGA-MD algorithms provide excellentfunction to improve the SAR image quality. Furthermore, we give a comparison of PGA-LS andPGA-MD, which makes it clear that the performance of PGA-MD is superior to PGA-LS. In the end,a new method utilizing PAST (Projection Approximation Subspace Tracking) is discussed whichsolves the problem of large computation amount of eigen-decomposition and ensures the estimateaccuracy simultaneously.Chapter3proposes a2D (Two-Dimensional) autofocus algorithm called the KA (Knowledge-Aided)2D autofocus algorithm. This method overcomes the shortcomings of blind estimate of phase errors ofthe algorithms discussed in chapter2. With the use of a priori knowledge of the phase structure, weobtain an analytic mapping relationship between the1D azimuth phase errors and the2D phase errorsunder the framework of PFA (Polar Format Algorithm). As a consequence, it just needs to estimate theazimuth phase errors and then by means of the mapping relationship the2D phase errors can be easilycomputed. In this chapter, this KA method is deduced in detail and via simulation and real dataprocessing this method is validated.Chapter4gives a discussion of the FPGA-based algorithm validation of PGA and the PAST-basedalgorithms using the KC705developing board with a Kintex7FPGA of Xilinx corporation. Since thesignal processing part needs to access the2D array in DDR3, a simplified interface is designed which supports four access modes which are read and write in sequential and transpose modes. Then statemachines of these two algorithms are designed with the one-iterative processing flow of PGA. Theresults show that with the clock of200MHz, it takes both of the two algorithms about0.5secondscomplete the processing of a SAR image of2048*2048samples. At last, the FPGA-processing resultsare analyzed in detail, which validates the FPGA implementations of autofocus algorithms. This worklays the foundation of the combination of the SAR autofocus algorithms and the basis imagingalgorithms in the future.Chapter5summarizes the work in this paper from two aspects, namely the research in SARautofocus algorithms and the FPGA implementations. What’s more, it sets the work in the future inalgorithm explorations and FPGA implementations. The authors will mainly get down to the2Dautofocus algorithms under other basic imaging algorithms based on the priori information for betterautofocusing guide. Besides, the authors will spare efforts in FPGA implementations to process largeramounts of SAR data and FPGA combination of autofocus and basic imaging algorithms to constructa complete imaging system.