【作者】 李新武；

【导师】 郭华东； 廖静娟； 王长林；

【作者基本信息】 中国科学院研究生院（遥感应用研究所） ， 地图学与地理信息系统， 2002， 博士

【摘要】 干涉雷达是基于雷达对同一区域采用不同视角成像得到的两幅雷达图像组合而建立的一门技术，主要利用了雷达信号的相位信息。它已广泛的应用于地形成图和地表形变探测，近几年来，它也被用作提取地表物理参数的一种重要的工具。极化干涉雷达技术是一种将极化和干涉雷达技术集成为一体的技术。雷达干涉主要对散射体的空间分布和高度敏感，雷达极化主要对散射体的形状和方向敏感，因此极化干涉雷达对有方向散射体的分布敏感，它已被用于反演地表植被参数，特别是森林的高度参数。极化干涉雷达是雷达遥感研究的前沿之一。本文基于ERS-1／2和SIR-C／X-SAR数据，对有关干涉和极化干涉信息提取及应用课题进行了广泛的研究，特别是首次对极化干涉雷达的基础理论进行了比较深入的论述，对极化干涉信息提取和应用进行了开拓性的发展，并取得了一些创新性成果，主要创新点包括： 1．在地表植被覆盖丰富的区域，常规干涉通过公共谱段滤波不能消除一个分辨率单元内由于散射中心高度差引起的去相干，而极化干涉相干最优过程能有效的消除这种谱去相干。因此在地表植被丰富的区域，用极化干涉提取的DEM应该具有比常规干涉DEM更好的精度。因此，本文基于SIR-CL波段单视复雷达数据，提出了一套完整的由极化干涉雷达数据生成数字高程模型的数据处理算法，并对其中的几个关键问题进行了研究。通过比较和分析，结果表明极化干涉DEM具有比常规干涉DEM更高的精度。还对极化干涉雷达相对于常规干涉雷达生成数字高程模型时不同的误差源进行了分析和讨论。 2．由于极化干涉相干最优过程提供了相位散射中心的最优分离，因此用极化干涉提取地表植被参数特别是森林的高度具有比极化或干涉更大的优势。但是该反演问题属于一个六维的非线性最优化问题，存在相干系数的低消光系数、低植被和高消光系数、高植被模糊性，地形相位的等值性。因此用一般的局部最优方法，它往往会与初始解的给定有很大的关系。因此，本文提出了基于模拟加温-退火算法的地表植被参数反演算法，改善了反演的精度，。 3．以二进小波理论、相位解绕理论和干涉图的相位特性为基础，本文提出了基于二进小波的雷达干涉条纹探测相位解绕算法。和其他基于条纹探测的相位解绕算法相比，它具有以下三个优点：(1)在多尺度边缘探测的同时能显著的压制噪声对边缘探测的影响；(2)利用多尺度局部模极大值的空间一致性条件，可以得到更为准确的相位突变边缘点；(3)利用梯度矢量流的活动等高线算法，可以连接由于被噪声严重破坏的边缘和区分相隔很近的边缘。 4．由于极化干涉提供了比极化或干涉更多的信息，在土地类型的识别和分类上具有比常规的干涉和极化更大的优势。本文基于SIR-C L波段雷达数据和研究区域的土地利用图，进行了地表地物的相干最优特性分析。并利用极化干涉最优相干系数、后向散射系数和极化熵进行地表土地类型的识别与分类实验研究，获得了好的分类结果。 5．初始基线的估计是干涉雷达数据处理的重点也是难点。本文基于二维FFT（傅立叶变换）的功率谱，实现了一个高精度估计初始基线垂直分量的算法。基于 SIR－C L波段单视复雷达数据，利用上面算法对初始基线的垂直分量进行了估计，同时结合从轨道数据模拟得到的基线平行分量进行了去平地相位实验，将去平地相位以后的结果与本区域的地形图相比较，具有相同的条纹变化规律。说明该算法估计的初始基线精度较高。 6．本文利用ERS＊ 重复轨道多时相干涉雷达数据，以昆仑山脉中段试验区为例，分析了该区域大部分地表无积雪和有积雪覆盖时地表各种地物的干涉相干特性，在此基础上利用相干系数、后向散射强度和后向散射强度的变化信息，进行了地表土地类型的识别和分类，获得了较好的分类结果。更多还原

【Abstract】 Synthetic aperture radar(SAR) interferometry is an established technique based on combining two SAR images of the same scene acquired from slightly different viewpoints. It is widely used for topographic mapping and surface change detection, moreover, in past few years, SAR interferometry has also been used to retrieve physical parameters of terrestrial surfaces. Polarimetric SAR interferometry is the technique that integrated the radar polarimetry and radar interferometry. Radar interferometry is sensitive primarily to the spatial distribution of vegetation, while polarimetry is sensitive primarily to the shape and orientation of vegetation scatters. Therefore polarimetric interferometry is sensitive to the distribution of orientated scatters. It has been used to invert the terrain surface vegetation parameters, especially for height of forest. It is a leading edge of radar remote sensing in recent years. In this dissertation, based on ERS-1/2 and SIR-C/X-SAR data, the project associating with SAR interferometry and polarimetric SAR interferometry information extraction and application have been extendedly studied. In particular, the basic theory of Polarimetric SAR Interferometry has been detailedly discussed, furthermore, the information extraction and application of Polarimetric SAR Interferometry are extendedly developed. Some new results has been obtained as follows:1. In the area covered by rich vegetation, conventional InSAR can not remove the decorrelation caused by height difference of scattering center inside the resolution elements by spectral filtering, but Ploarimetric SAR Interferometry can effectively remove this kind of spectral decorrelation. The DEM obtained from Polarimetric SAR Interferometry should have higher accuracy than that obtained from conventional InSAR. Based on SIR-C L band complex single look data, one integrity data processing algorithm that generated DEM from Pol-InSAR data has been presented, furthermore, several key problems have been studied. The results indicate that the DEM obtained from Polarimetric SAR Interferometry have higher accuracy than that obtained from InSAR. Also, the error resources that different from the conventional InSAR have been discussed.2. Being the coherent optimization of polarimetric SAR interferometry provided the optimal separation of phase center, it has more advantages to retrieve the terrain surface vegetation parametes than that of conventional InSAR or PolSAR. But it belong to six dimensional inversion problem, It has the ambiguity of low extinction, low height and high extinction, high height of vegetation in coherence and equivalence of topography phase. So using the conventional local optimization algorithm, it often get stuck in a local minimum when algorithm is restarted with different initial guess value. At this dissertation, the simulated annealing algorithm, which can give global solution of problem, has been implemented in order to improve the inversion accuracy.3. Based on dyadic wavelets theory, phase unwrapping theory and the phase characteristics of interferogram. One phase unwrapping algorithm that based on interferometric fringe detection is presented, compared with the other phase unwrapping method that based on the fringe line detection, It have three advantages. First, it can suppress significantly the effects of noise when multi-scale edge detection is implemented. Second, using the space consistency condition of multi-scale local module maximum, the more accurate phase sharp variation edge can be obtained, Third, using the active contour algorithm - GVF (Gradient Vector Flow) snake, it can connect the fringe line when the phase sharp variation edge is seriously corrupted by the noise or the fringe line are closely spaced.4. Being the polarimetric SAR interferometry provided more information than that conventional InSAR or PolSAR, It is advantageous for identification and classification of land cover. Utilizing the fully polarimetric SIR-C L band data and land-use map of study area, the更多还原