【作者】 曲小宁；

【导师】 冯大政；

【作者基本信息】 西安电子科技大学 ， 信号与信息处理， 2013， 博士

【摘要】 干涉合成孔径雷达(InSAR)是在合成孔径雷达(SAR)的基础上，雷达成像技术的新发展, InSAR能够根据两幅SAR图像对应像素点之间的绝对相位差所反映的距离差来获得目标高度,具有全天候、全天时、高效率获取目标区域数字高程地图(DEM)的能力，在军事、国民经济建设和科学研究中，都有着极其广泛的应用，因此对InSAR技术的研究，在国内外受到高度重视。论文以星载InSAR干涉成像技术为研究对象，在详细地介绍了InSAR技术发展历程、应用以及有关的InSAR成像原理基础上，对InSAR干涉成像处理流程中各个阶段做了系统地研究。在InSAR信号处理的过程中，由于相位图像的相位差被限制在(-π,π]之间,因此模糊相位的展开是干涉合成孔径雷达信号处理的关键步骤之一。但由于噪声、欠采样等因素的影响,精确的相位展开变得非常困难，因此如何获得高精度的展开相位,一直是InSAR数据处理中的热门研究问题。本论文对相位展开算法进行了深入研究，并在单基线相位展开基础上对多基线InSAR相位展开技术进行了初探，以提高干涉数据的利用率和获取DEM的精度。具体工作内容如下：1.对相位展开方法中基于路径跟踪的相位展开方法进行了研究，在Goldstein枝切法的基础上，利用旅行商问题中求解最短路径的思想，提出一种利用改进的遗传算法建立连接正负残差点的最短枝切线进行相位展开的算法，这种算法相较Goldstein枝切法能够有效缩短建立的枝切线长度，避免了因枝切线包围造成孤立区间无法解缠的“孤岛现象”，有效的提高了枝切法在残差点密集区域的相位展开精度。2.对相位展开方法中的全局法进行了研究，在最小二乘法的基础之上提出了一种利用无网格法的相位展开法，该算法首先通过径向点插值无网格法计算出相位在划定的支持域内的近似函数，然后在最小二乘意义下通过计算待展开相位点在支持域范围内的局部最佳近似值进行相位展开。由于结合了最小二乘法和无网格法的优点，该方法对于干涉图中质量较低区域同样可以进行有效的相位展开，同时在很大程度上防止了误差的传递。3.提出了一种质量引导法和B-样条曲面拟合相结合的相位展开算法，利用质量图将干涉相位图划分为高质量区和低质量区，首先利用质量引导法求得高质量区的展开相位，然后根据已获得的展开相位，采用B-样条曲面拟合的方法对低质量区的相位进行计算求得最终的展开相位。该方法通过划分区域，由高质量区向低质量区进行相位展开，可以有效的减少误差的传递同时展开低质量区的相位时充分利用了已获得的高可靠性展开相位，有效的提高了相位展开的稳健性。4.多基线干涉合成孔径雷达是传统单基线干涉合成孔径雷达的扩展，它优点在于能够利用长短基线各自的优点，同时可以根据长短基线间的关系更加准确的估计出绝对相位。论文对多基线InSAR系统进行了初步研究，提出了一种广义稳健的中国余数定理来进行相位展开的方法，该方法根据多条长度具有参差关系的基线测量得到的干涉相位，通过求解同余方程组来进行相位展开，同时分析了相位噪声对相位展开所造成的影响，给出了正确解模糊允许的最大相位噪声。更多还原

【Abstract】 The Interferometric synthetic aperture radar (InSAR) is a new radar imagingtechnology that based on synthetic aperture radar (SAR). InSAR can obtain the heightof the targts from phase differences of the two synthetic aperture radar (SAR) images,and it can get digital elevation model (DEM) of the nominated target in all weather andall time effectively. In military, the construction of national economy and scientificresearch field, InSAR has a wide range of applications, thus the research of the InSARhas been highly regarded both here and abroad.The interference imaging technology of spaceborne-InSAR as the research objectdeveloped here, the development of the InSAR technology, applications and InSARImaging principle are introduced, each step in interference imaging technology of theInSAR have been researched deeply.Since the differences of measured phase is wrapped into (-π,π], the phaseunwrapping is a key steps in interference synthetic aperture radar(InSAR) signalprocessing. However, since the presence of different error such as noise or undersampled, the real phase cannot be retrieved straightforwardly, there need to developphase unwrapping algorithms to deal with this problem. So in this thesis phaseunwrapping problem is developed, and to improve the utilization of Interferometric dataand the precision of DEM, multiple baseline InSAR phase unwrapping technique isdiscussed that based on single baseline phase unwrapping. The content of the thesis is asfollows:1. How to establish branch-cut is very important to Path following method, the shorterthe length of the branch-cut, the better the result of phase unwrapping. Therefore, theimproved genetic algorithms for optimizing the length of the branch-cut is presentedbased on Goldstein algorithm, which has already been successfully applied to find theshortest path in the traveling salesman problem (TSP) theory. The method can minimizethe length of the branch-cut and avoid the "isolated island phenomenon" of unwrappingprocess effectively, and the precision of the phase unwrapping has been improved inhigh residue regions.2. The global method of phase unwrapping developed here, a robust method forunwrapping2D phase by using mesh-less method under least-squares theory isproposed. Mesh-less function is obtained by radial point interpolation method and phaseunwrapping process is implemented by estimating the local optimal approximate value under an efficient least-squares solution. As the proposed method takes advantage bothfrom least-squares and mesh-less method, it cannot only unwrap the unreliable areascorrectly, but also can forbid error accumulate to the whole image.3. A hybrid phase unwrapping method that combines the quality guided method andB-spline surface-fitting method is proposed in this thesis. The proposed method utilizesdifferent methods at different quality regions to realize the phase unwrapping, it usesquality guided method to unwrap the phases in high quality regions, and uses B-splinesurface-fitting method to obtain the surface of the unwrapped phase in low qualityregions. The wrapped phase can be unwrapped from high quality region to low qualityregion which can reduce the error accumulate and improve the robustness of phaseunwrapping.4. Multi-Baseline InSAR is expand of the tradition Single-SAR, according to therelationship of the long and the short baseline, the wrapped phase can be unwrappedsimply. The Multi-Baseline InSAR developed here, and a phase unwrapping methodbased on a generalized robust Chinese Remainder Theorem is proposed with detailderivation, this method can use the relationship of the long and the short baseline toobtain the nterferometric phase, realize the phase unwrapping by solving the congruenceequation. The impact of the phase noise to the unwrapping was analyzed; the range ofthe maximum tolerable phase noise resolve the correct ambiguity was given.更多还原