【作者】 何儒云；

【导师】 王耀南；

【作者基本信息】 湖南大学 ， 控制理论与控制工程， 2007， 博士

【摘要】 合成孔径雷达干涉测量(InSAR)技术是近二十年发展起来的一种先进的空间观测技术,论文以星载InSAR干涉成像技术为研究对象,在简单介绍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,并结合实例对其中关键处理技术包括SAR图像配准、干涉图滤波、干涉图相位解缠的原理和算法做了深入地分析和研究,利用真实ERS-1/2 SAR图像数据干涉测量生成相对数字高程图,利用ENVISAT ASAR图像数据生成同一地区的干涉相位图。论文针对SAR图像配准过程中概略配准的必要性,依据地理坐标参数(经纬度)和SAR图像坐标之间的对应关系,提出了一种基于地理坐标参数的SAR图像自动概略配准法,拟合地理坐标和图像坐标之间的变化,实现概略配准,算法计算速度快,精度较高。在概略配准的基础上,分析了精确配准的各种方法和干涉相位的统计特性、估计及评价。论文在分析了干涉图去平地效应原理与方法的同时,讨论了干涉图相位噪声特性和干涉图滤波方法,并利用小波分析的多分辨率特性和高频部分系数的方向特性,提出了一种基于小波变换的InSAR干涉图滤波方法,在对干涉图小波分解的基础上,对高频部分系数用对应方向的方向模板平滑处理后做中值滤波,使干涉图中的残差点数目大大减少,同时很好地保持了图像的边缘特征。论文分析比较了相位解缠中的路径跟踪法和最小二乘法,针对相位解缠中相位噪声引起的误解缠问题,利用支持向量机强大的分类能力,提出了一种基于支持向量机的InSAR干涉图相位解缠法,根据像元的相位属性对像元分类,对不同类的像元用不同的方法解缠,算法尤其适合解缠具有统计特性的真实干涉图;利用Mumford-Shah提出的最小分割问题的“多层”实现模型,提出了一种基于多层Mumford-Shah模型的InSAR干涉图相位解缠方法,采用“水平集逐层迭代算法”对干涉图进行多层分割以间接确定不同质量特性的干涉图子连通区域,在此基础上从高质量区域向低质量区域对干涉图中像元逐步进行相位解缠,算法先隔开噪声密集的低质量区域,对干涉图进行解缠,提高了解缠的精度;利用枝切法中隔离残差点的思想,提出了一种基于最短距离优先原则的InSAR干涉图区域分割相位解缠法,采用全局性最短距离优先原则连接残差点,把干涉图分割为非残差区域和残差区域,分别用不同的方法解缠,在保证高质量的非残差区域快速准确解缠的同时,使更多的低质量区域像元准确解缠。论文利用三组ERS-1/2串行的SAR图像数据对,进行干涉成像处理实验,并根据干涉处理中得到各组数据的不同特点,使用不同的干涉成像处理方法,成功地生成了相关系数图、干涉相位图和成像目标区域的相对数字高程图,从而进一步检验了论文给出方法的有效性。同时,利用由ENVISAT卫星ASAR对BAM地区地震前后成像得到的多幅SAR单视复图像数据,选用一组降轨图像数据和一组升轨图像数据分别做干涉成像处理,生成干涉相位图,可以进一步利用合成孔径雷达差分干涉测量技术和有关地震知识研究地震地表形变。更多还原

【Abstract】 Interferometric Synthetic Aperture Radar ( InSAR ) is an advanced space observation technique that have been developed in the last two decades or so, this dissertation studies the spaceborne InSAR imaging technique, after the development process and application of InSAR and the related InSAR imaging theory have been simply introduced, the InSAR process flow is systematically studied, then the key techniques such as image registration, noise filtering and phase unwrapping with related examples are fully analyzed and discussed,and the relative digital elevation models are generated using the real ERS-1/2 SAR data,and the SAR interferograms are generated using the ENVISAT ASAR data of the same imaging targets.To the requirement of coarse registration in the process of SAR image registration, an automatic coarse SAR image registration algorithm is presented based on the geographical coordinates parameters according to the corresponding relation between the geographical coordinates parameter and image coordinates, the method is implemented by fitting the variance of the corresponding geographical coordinates along with the pixels’image coordinates, it is fast and with high precision. Based on the coarse registration, several algorithms of fine registration and the characteristics and estimation and evaluation of the interferometric phase are analyzed.After the principle and methods of flat earth removal in SAR interferogram are investigated, the characteristics of phase noise and speckle reduction method are discussed, an InSAR interferogram filtering algorithm based on the wavelet transform is presented utilizing the multiresolution characteristics and the detail coefficient orientation characteristics of wavelet analysis, after the SAR interferogram is decomposed using wavelet transform, the method smoothes the coefficient using different mode corresponding to its orientation, then filters the coefficient using median filtering. The method decreases the residues, and with good edge preservation.The path followed phase unwrapping algorithm and least square phase unwrapping algorithms are discussed and compared, aiming at the error caused by noise, an InSAR phase unwrapping algorithm based on support vector machines is presented using its powerful classification ability, the pixels are classified using the phase-related information, the different type of pixels are unwrapped by different methods, it is an effective method especially for the real SAR interferogram with phase statistics. An InSAR phase unwrapping method based on Hierarchical Mumford-Shah functional Model is presented, the SAR interferogram is segmented into many sub-connective regions with different quality properties by an iterative tier-by-tier level set algorithm, then these regions are unwrapped region by region from the high-quality ones to low-quality ones, thus the method can unwrap the pixels after the low-quality regions have been isolated, which improves the accuracy of phase unwrapping. A region-cutting InSAR phase unwrapping algorithm based on the Shortest Distance First Principle utilizing the idea of isolated the residues, which adopts the global method to connect the residues,the SAR interferogram is cut into non-residue regions and residue regions, different methods are adopted to unwrap them, which guarantees that the pixels of non-residue regions can be unwrapped with accuracy, moreover, much more pixels of residue regions can be correctly unwrapped.The spaceborne InSAR experiments are carried out using the three groups ERS-1/2 SAR tandem data, the different processing algorithms of different group are adopted to these groups according to the group’s own characteristics, the correlation map, SAR interferogram and the relative digital elevation models of imaging target are obtained. In addition, several SAR interferograms are generated using one group of the descending pass SLC images and one group of the ascending pass SLC images of the BAM earthquake obtained by ENVISAT ASAR, then the coseismic deformation can be studied using D-InSAR technique and related earthquake knowledge.更多还原