SAR自聚焦算法研究及其FPGA实现

发布时间：2018-02-04 08:39

本文关键词： SAR 自聚焦算法多子孔径自聚焦两维自聚焦 FPGA 实时处理　出处：《南京航空航天大学》2014年硕士论文　论文类型：学位论文

【摘要】：自聚焦算法是SAR（SyntheticAperture Radar，合成孔径雷达）成像处理中的一个重要步骤，，它研究的是估计经运动补偿后的残留相位误差并进行补偿以提高SAR图像的质量，是对运动测量单元的补充。因此，本文对自聚焦算法及其FPGA实现进行了一系列的探究。第一章介绍了本文研究的背景及意义，对自聚焦算法的发展及其基于FPGA（FieldProgrammable Gate Array，可编程逻辑门阵列）的硬件实现进行了回顾与总结，比较了常用的各种自聚焦算法的优缺点。第二章详细分析了几种估计一维方位相位误差的自聚焦算法。在这一章里，首先给出了PGA算法的处理流程，分析了估计的性能；针对PGA算法收敛速度慢的缺陷，介绍了QPGA（QualityPGA）算法的原理。然后以PGA算法为基础，讨论了处理长CPI（Coherent Processing Interval，相干处理间隔）图像的几种算法：PCA（Phase CurvatureAutofocus，相位曲率自聚焦），PGA-LS，PGA-MD，给出了算法的推导过程，算法性能分析及实测数据的处理。处理结果表明，全孔径PGA与PCA算法处理长CPI散焦问题时是失败的，PGA-LS和PGA-MD能够有效的改善原始SAR图像散焦效果。进一步，我们对PGA-LS与PGA-MD算法进行了比较，分析结果表明PGA-MD算法性能优于PGA-LS。最后介绍了基于PAST（Projection Approximation SubspaceTracking，投影近似子孔径跟踪技术）的自聚焦算法。该算法克服了基于特征分解方法的自聚焦算法计算量大的问题，且估计的精度高鲁棒性好。在第三章中我们介绍了一种基于先验知识的两维自聚焦算法。这一方法克服了第二章中几种算法对相位误差直接进行估计而忽略究竟采用何种算法得到SAR图像的缺陷，充分利用了先验的信息，获得了在PFA（Polar Format Algorithm，极坐标格式算法）框架下两维相位误差与一维方位向相位误差的解析关系。因此，只需估计出方位向的相位误差即可完成两维相位误差的估计。我们给出了这种解析关系的详细推导，并作了仿真验证及实测数据的处理，处理的结果表明了本算法的有效性。第四章讨论了利用Xilinx公司KC705开发板（Kintex7系列FPGA）实现PGA与PAST两种自聚焦算法。由于FPGA需要频繁的从DDR3读取数据，为更方便的访问DDR3中的二维SAR数组，设计了一种访问DDR3的四种工作（转置读写）模式的状态机，简化了接口的使用。分别设计了两种算法的状态机及PGA算法的一次迭代处理流程图，两种算法在200MHz的系统时钟下以0.5s左右的时间处理了一幅2048*2048大小的SAR图像，并对FPGA处理结果进行了详细的分析，分析结果表明FPGA实现自聚焦算法的可行性和有效性。这为今后将自聚焦算法与基本成像算法结合起来形成一个完整的SAR实时处理系统奠定了基础。第五章对本文的两大部分内容，即SAR自聚焦算法研究与FPGA的硬件实现作了一个总结。本章还讨论了今后的工作安排与展望，指出了未来的任务有两个：一方面将继续探讨其他的自聚焦算法，尤其是在除PFA外的其他成像算法下的两维自聚焦算法，为SAR成像处理提供更好的理论依据；二是将利用FPGA芯片间的相互通信实现数据量更大的聚焦处理并将自聚焦算法与基本成像算法结合起来以实现一个完整的SAR成像处理系统。
[Abstract]:Autofocus algorithm is SAR (SyntheticAperture Radar, synthetic aperture radar) is an important step in image processing, it is estimated by the residual phase error after motion compensation and compensation to improve the quality of SAR images, is a supplement to the motion measurement unit. Therefore, this paper explores a series of implementation autofocus algorithm and FPGA.
The first chapter introduces the background and significance of this paper, reviews the development of self focusing algorithm and its hardware implementation based on FPGA (FieldProgrammable Gate Array, programmable logic gate array), and compares the advantages and disadvantages of various self focusing algorithms.
The second chapter analyzes several estimation autofocus algorithm for one-dimensional azimuth phase errors. In this chapter, firstly, the process of PGA algorithm, the estimation performance is analyzed; according to the defects of PGA algorithm with slow convergence, introduces QPGA (QualityPGA) algorithm principle. Then based on the PGA algorithm, is discussed the length of CPI (Coherent Processing Interval, coherent processing interval) several algorithms of image: PCA (Phase CurvatureAutofocus, PGA-LS, phase curvature autofocus), PGA-MD, derivation of the algorithm is given, processing algorithm and performance analysis of the measured data. The results show that the full aperture PGA and PCA algorithm for defocus length CPI the problem is a failure, PGA-LS and PGA-MD can effectively improve the original SAR image defocusing effect. Further, we compare PGA-LS with PGA-MD algorithm, and the results show that the PGA-MD algorithm is of better performance At the end of PGA-LS., a self focusing algorithm based on PAST (Projection Approximation SubspaceTracking, projection approximation subaperture tracking technology) is introduced. The algorithm overcomes the problem of large computation based on eigendecomposition algorithm, and has high accuracy and robustness.
In the third chapter, we introduce a two dimensional autofocus algorithm based on prior knowledge. This method overcomes several algorithms in chapter second of the phase error to be estimated directly and ignore what exactly SAR image defects which kind of algorithm, make full use of the prior information obtained in PFA (Polar Format Algorithm, a the algorithm) under the framework of two dimensional phase error and one-dimensional azimuth analytical relationship of phase error. Therefore, only need to estimate the phase error estimation of azimuth can be completed two dimensional phase error is derived in detail. We give the analytic relationship, and makes the processing simulation and measured data, processing the results showed that this algorithm is effective.
The fourth chapter discusses the use of Xilinx KC705 development board (Kintex7 series FPGA) PGA and PAST two kinds of autofocus algorithm. Because FPGA requires frequent read data from DDR3, two-dimensional SAR array is more convenient access to the DDR3, designed four kinds of a DDR3 access (read and write to the state) machine mode, simplify the use of the interface are designed. The state machine of PGA algorithm and two algorithms of an iterative processing flow chart, two kinds of algorithm in the 200MHz system clock to 0.5s around the time of processing SAR images of 2048*2048 and FPGA on the size of the processing results are analyzed in detail the results show that the FPGA, the feasibility and effectiveness of the autofocus algorithm. This is the future of the autofocus algorithm with basic imaging algorithm are combined to form a complete SAR real-time processing system of the foundation.
The fifth chapter two most of the contents of this article, namely SAR and FPGA study on self focusing algorithm hardware implementation has made a summary. This chapter also discusses the future work arrangements and prospects, and points out that the future has two tasks: on the one hand will continue to explore other autofocus algorithm, especially in the other two dimensions in addition to PFA imaging algorithm under the autofocus algorithm, provides a better theoretical basis for SAR imaging; two is the use of mutual communication between the FPGA chips to realize data volume focusing more and the autofocus algorithm with basic imaging algorithms are combined to achieve a complete SAR imaging system.

【学位授予单位】：南京航空航天大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.52

【参考文献】