【作者】 李延伟；

【导师】 梁甸农；

【作者基本信息】 国防科学技术大学 ， 信息与通信工程， 2010， 博士

【摘要】 极化干涉合成孔径雷达(Synthetic Aperture Radar,SAR)是将极化SAR和干涉SAR有机结合的新型雷达体制,它利用多幅极化SAR干涉图像进行信息提取,在植被参数反演和地面运动目标检测领域取得了广泛应用。极化干涉SAR复相关系数模型是极化干涉SAR的基础,决定着极化干涉SAR信息提取精度,而极化干涉SAR复相关系数模型依赖于干涉雷达的收发模式。最初的极化干涉SAR系统都是两部雷达各自收、发的单基地极化干涉SAR系统,随着一发多收技术的发展,极化干涉SAR系统正由单基地体制向着双基地(本文中的双基地极化干涉SAR专指一部雷达自发自收、而另一部雷达被动接收的干涉系统)体制转变,双基地极化干涉SAR是未来雷达技术的发展趋势,是当前雷达遥感领域的研究热点。本文以单/双基地极化干涉SAR植被参数反演和运动目标检测为研究背景,对极化干涉SAR协方差矩阵的高精度估计、单/双基地两层植被极化干涉建模及其植被参数反演新方法、单/双基地三层植被极化干涉建模及其植被参数反演新方法、单/双基地沿航向极化干涉SAR-GMTI性能提高方法及系统参数优化方法等关键问题进行了研究。论文的主要工作包括:第二章主要介绍单/双基地极化干涉SAR基础。首先,建立了单/双基地全极化线性调频信号的发射、接收和全极化SAR图像信号模型;然后,基于全极化SAR图像信号模型,建立了各类目标通用的极化干涉SAR信号模型;最后,基于极化干涉SAR植被参数反演信号模型,阐述了植被参数反演的基本原理,分析了体相关系数估计误差对经典三阶段植被参数反演方法性能的影响。第三章研究极化干涉SAR协方差矩阵高精度估计方法。提出了基于区域增长寻窗和相位补偿的极化干涉SAR协方差矩阵迭代估计方法。该方法首先挑选干涉条纹质量最好的极化单视复相位图像进行旋滤波,获得干涉相位估计值,并利用该干涉相位估计值对干涉矩阵进行相位补偿,消除相位非平稳性的不利影响;然后,利用全极化强度图像进行区域增长寻窗,保证多视窗口里的像素满足散射同质性;最后,对于满足散射同质性和相位平稳性的窗口,提出了基于Hermitian乘积模型的协方差矩阵迭代估计方法。该方法不仅可抑制散射异质性和相位非平稳性导致的协方差矩阵估计偏差,还可减小低相干性导致的估计偏差,具有较高的极化干涉SAR协方差矩阵估计精度。实验结果验证了本文方法的优越性。第四章研究两层植被单/双基地极化干涉模型及其植被参数反演方法。建立了两层植被单/双基地极化干涉SAR复相关系数模型,重点分析了收发模式和散射机理对极化干涉SAR复相关系数分布特征的影响,进而研究了收发模式和散射机理对三阶段植被参数反演方法精度的影响,发现了新模型下三阶段植被参数反演方法精度不高的原因。提出了基于BP神经网络的极化干涉SAR植被参数反演方法,该方法利用BP神经网络直接模拟极化干涉复相关系数与植被参数之间的非线性关系,避免了三阶段植被参数反演方法面临的地面干涉相位模糊和体相关系数估计误差的不利影响,同时,由于该方法直接拟合复相关系数与植被参数之间的非线性关系,不受极化复相关系数分布特征的影响,因此,该方法同样适合于双基地极化干涉SAR植被参数反演。提出了基于Freeman分解的极化干涉SAR植被参数反演方法,该方法利用极化协方差矩阵、极化干涉矩阵与Freeman分解之间的关系,把植被参数反演问题转化为非线性优化问题,进行植被参数反演,具有思路直观、反演精度高、速度快等特点。更重要的是该方法不受极化干涉SAR复相关系数分布特征的限制,即使对于双基地极化干涉SAR系统,该方法仍然适用,实验结果验证了新方法的有效性。第五章研究单/双基地三层植被极化干涉模型及其植被参数反演方法。建立了三层植被单/双基极化干涉SAR复相关系数新模型,分析了三层植被单/双基极化干涉SAR复相关系数分布特征,讨论了新模型下三阶段植被参数反演方法的局限性,提出了基于多基线极化干涉SAR的三层植被参数反演方法。三层植被模型是两层植被模型的扩展,基于多基线极化干涉SAR的三层植被参数反演方法,不仅能够提高植被参数反演精度,更重要的是,还能够获得更多的植被参数信息,实验结果验证了新模型的正确性和新方法的有效性。第六章研究单/双基地极化SAR-ATI运动目标检测及系统参数优化设计。针对单基地极化沿航向干涉SAR系统,提出了极化虚拟多基线的概念和构造方法,分析了基于极化虚拟多基线的运动目标检测性能,比较了极化虚拟多基线相对于单极化沿航向干涉SAR系统的运动目标检测优势。建立了以最大化不模糊可测速区间长度比为目标函数的系统优化模型,优化模型的自变量是实际基线长度和脉冲重复周期,经过模型优化处理得到的实际基线长度和脉冲重复周期的组合能够将极化信息在提高运动目标检测能力方面的优势最大化。针对乒乓模式的全极化沿航向干涉SAR系统,研究了其极化虚拟多基线的构成及特征,分析了新模式下极化沿航向干涉SAR系统的运动目标检测潜能,并以TanDEM-X和TanDEM-L系统为参考,分析了基于极化虚拟多基线的运动目标检测潜能。更多还原

【Abstract】 Polarimetric SAR Interferometry (Pol-InSAR) is an advanced SAR system, which combines both the advantages of the Polarimetric SAR and those of the interferometric SAR. It can get the interferometric information of many different polarimetric SAR, so it has been widely used in the domains of vegetation parameters inversion and moving target detection. The model of the Pol-InSAR complex correlation coefficient is the base of the information extraction and decides the precision of the information extraction. Especially, the study on the Pol-InSAR system has been changed from the mono-static mode to the bi-static mode. In this dissertation, the mono-static Pol-InSAR system means that both radars transmit the radar signal and receive its signal at the same time. The bi-static Pol-InSAR system means that only one of the radars transmits the signal and both radars receive the signal. The bi-static Pol-InSAR system is a developing trend of the radar and is a study hotspot in the current domain of radar remoting sensing.This dissertation focuses on the vegetation parameters inversion and the moving target detection by the mono/bi -static Pol-InSAR. The research topics include the highly precise estimation of the Pol-InSAR covariance matrix, the modeling of the mono/bi -static Pol-InSAR complex correlation coefficient of the two-layer vegetation structure and the new parameter inversion method to it, the modeling of the mono/bi -static Pol-InSAR complex correlation coefficient of the three-layer vegetation structure and the new parameter inversion method to it, the method to improve the detection performance of the moving target by the fully-polarimetric along-track interferometry SAR system and the optimization method to the along-track Pol-InSAR system. The main work of this dissertation includes:Chapter 2 introduces the base of the Pol-InSAR. The receive-transit signal model of the mono/bi -static fully-polarimetric LFM signal and the mono/bi- static fully-polarimetric SAR image signal model are established firstly. Then, based on the fully-polarimetric SAR image signal model, the generalized Pol-InSAR signal model is proposed. Finally, based on the signal model of the vegetation parameters inversion, the basic principle of vegetation parameters inversion is expatiated and the effect to the performance of the three-stage inversion method is analyzed.In chapter 3, the method to the highly precise estimation of the Pol-InSAR covari ance matrix is investigated. A novel method based on the adaptive neighborhood region-growing principle and the compensated phase is proposed. Firstly, the bad effect by the phase nonstationarity is eliminated by the way in which the interferometric matrix is compensated with the interferometric phase estimated from the polarimetric interferogram with the best interferometric performance. Then, the region-growing principle is used to the fully-polarimetric SAR intensity images to get a window, in which the pixels satisfy the local scattering stationarity hypothesis. To the window satisfying the local phase stationarity hypothesis and the local scattering stationarity hypothesis, a novel iterative method based on the Hermitian product model is proposed. This new method could overcome the estimation errors caused by the nonstationarities both in the phase and in the scattering intensity, furthermore, it takes into account the characteristics of the Hermitian model, so it has a high estimation precision. Experimental results validate the superiority of this method.In chapter 4, the two-layer vegetation Pol-InSAR correlation coefficient model to the bi-static Pol-InSAR system and the vegetation parameters inversion method to it are studied. A new model to the two-layer vegetation complex correlation coefficient of the bi-static Pol-InSAR system is established. Based on this new model, the effect of scattering components and the receive-transit mode of the interferometric radar to the complex correlation coefficient distribution is analyzed, furthermore, the effect of scattering components and the receive-transit mode of the interferometric radar to the inversion precision of three-stage method is studied. The reason why the inversion precision of the three-stage method is low is discovered. A new vegetation parameter inversion method based on the BP neural network is proposed. This new method uses the BP neural network to fit the nonlinear mapping relationship between the complex polarimetric correlation coefficients and the vegetation parameters. It not only reduces the error caused by the error in the estimations of the ground interferometric phase and the volume correlation coefficient, but also reduces the error caused by the scattering model error. Because this method is not affected by the distribution characteristics of complex correlation, this new method is suitable to vegetation parameter inversion by bi-static Pol-InSAR working in one-transmit bi-receive mode. A Novel vegetation parameter inversion method based on the Freeman decomposition is proposed. Based on the relationship among the polarimetric covariance matrix, the polarimetric interfero- metric matrix and the Freeman decomposition, the new method models the problem of the vegetation parameter inversion as a nonlinear optimization problem, the variables of which are the vegetation parameters. Compared to three-stage inversion process, this new method supplies a new approach to the inversion of the vegetation parameters and has higher precision and quicker computing velocity. This new method doesn’t suffer from the limitation of the distribution characteristics of the Pol-InSAR correlation coefficient, so this method could be used to the vegetation parameter inversion by bi- static Pol-InSAR working in the one-transmit bi-receive mode too. Experimental results validate the correctness of the new model and the superiority of this method.In chapter 5, the three-layer vegetation Pol-InSAR correlation coefficient model to bi-static Pol-InSAR system and the vegetation parameter inversion method to it are studied. A new model to the three-layer vegetation complex correlation coefficient of the bi-static Pol-InSAR system is established. Based on this new model, the distribution characteristics of its Pol-InSAR correlation coefficient is analyzed, the limitation of the three-stage inversion method is discussed and a new vegetation parameter inversion method based on the dual-baseline Pol-InSAR is proposed. Three-layer vegetation construction is the extension of the two-layer vegetation construction. Compared to the single-baseline Pol-InSAR, this new method could improve the parameter inversion accuracy and estimate more vegetation parameters. Experimental results validate the correctness of the new model and the superiority of this new method.In chapter 6, a method to improve the moving target detection performance of the fully-polarimetric along-track interferometric SAR system and the optimization design method to the fully-polarimetric along-track interferometric SAR System are studied. To the fully-polarimetric along-track interferometric SAR System working in the pursuit mono-static mode, the concept and the construction method of the polarimetric virtual multi-baseline are proposed. The potential of the moving target detection by the polarimetric virtual multi-baseline is analyzed and advantages of the fully-polarimetric along-track interferometric SAR System to the the single polarimetric along-track interferometric SAR System is compared. An optimization model in which the object function is the proportion of the detectable velocity length of the two interferometric SAR systems is constructed. The variables of the optimization model include the real length of the baseline and the pulse repetition period. The optimization results of the variables maximize the superiority of the polarimetric information in the improvement of the moving target detection performance. To the fully-polarimetric along-track interferometric SAR system working in the alternating bi-static mode, the construction and the characteristics of the polarimetric virtual multi-baseline are studied. The potential of the moving target detection by the polarimetric virtual multi-baseline of this new mode is analyzed. Especially, the potential of the moving target detection by the polarimetric virtual multi-baseline of this new mode is analyzed based on the parameters of the TanDEM-X and the TanDEM-L system.更多还原