【作者】 彭曙蓉；

【导师】 王耀南；

【作者基本信息】 湖南大学 ， 模式识别与智能系统， 2009， 博士

【摘要】 合成孔径雷达干涉测量(InSAR)技术是近二十多年发展起来的一种先进的空间观测技术,它将传统的合成孔径雷达技术与干涉测量技术结合起来,由双天线模式或单天线重复轨道模式对同一地区雷达成像生成两幅SAR复图像,利用这两幅复图像的干涉相位信息提取地表的三维信息和高程变化信息,获取全球高精度的地形信息。它具有高分辨率、全天候全天时工作、成像范围宽和应用范围广等特点。而差分InSAR(D-InSAR)技术是在InSAR基础上发展起来的,可以监测地球陆地表面和冰雪表面的微小变化,从而获得高精度、高可靠性的地表变化信息。这项技术可用于研究过去长时间无法到达的冰川和冰缘的变化情况,也可用于一些灾害性地表形变的探测,如地震、火山爆发、滑坡、洪水等,尤其是其在军事方面的应用引起了各国的关注。因而InSAR已成为卫星遥感研究的热点。论文系统介绍了InSAR技术的发展历程、应用现状、几何成像原理及数据处理流程;对其中的关键处理技术包括SAR图像配准、干涉图去平地效应和去噪、干涉图相位解缠的原理和算法做了深入的讨论和分析;研究了InSAR与D-InSAR技术在地球科学中的应用,并给出了应用实例。用于干涉测量的两幅SAR复图像必须先配准。为获取高精度的测量结果,论文提出一种包括概略配准、像素级初配准和亚像素级精配准三个环节的逐级提高精度的InSAR复图像配准方法。该方法利用EnView软件提供的经度和纬度实现概略配准;采用基于多边形结构和边缘提取的像素级初匹配方法大大提高匹配速度;考虑以往的匹配算法都是基于能量图进行并且忽略了SAR复图像的整体性,采用基于整体松弛匹配和基于相位差分的最小二乘匹配相结合的亚像素级精配准算法减少误匹配率,提高匹配精度。配准后生成的InSAR干涉图相位信息包括地形信息和平地效应信息,为获取精确地形信息,必须去除平地效应信息。针对常规去平地效应方法中参数难以获取及精度不够高的现状,提出了基于线性插值的频谱偏移去平地效应方法,该算法在体现频移去平地效应的优势的基础上,通过插值提高频移的分辨率。实验结果表明该算法容易实现,精度也大大提高。具有各向异性特征的InSAR干涉图中含有各种噪声,常规方法往往采用矩形滤波窗口去噪,这样做的结果是在得到滤波效果的同时模糊了相位信息,损害了干涉图的条纹信息。论文提出一种基于干涉条纹中心的InSAR干涉图滤波算法,该算法采用沿干涉条纹中心线获取的等相位线滤波窗口滤波,符合干涉图的各向异向特征;考虑到获取条纹中心线的复杂性,使用CNN(细胞神经网络)的并行算法来提高获取条纹中心线的速度;鉴于干涉相位图存在2π跳变线,最后的滤波采用正弦图(或余弦图)均值滤波算法。用该方法实现了两幅实际图像的滤波,实验结果的视觉特征良好、残差点减少、边缘保持能力增强,表明该算法具有高精度的滤波效果。常用的相位解缠算法分为两种类型:基于路径跟踪的相位解缠方法和基于最小二乘法的相位解缠方法,这两类方法面临精确性和计算速度不能兼顾的问题。对此本文提出一种基于Delaunay三角网最小费用流算法的分区域干涉图相位解缠方法。该方法采用残差点距离原则利用残差点信息将干涉相位图分为高质量相位区域和低质量相位区域;考虑到残差点分布的不规则性和最小费用流算法的成熟性,利用基于Delaunay三角剖分的最小费用流解缠算法和区域生长解缠算法的优点,分别对InSAR干涉图相位高质量区域和低质量区域解缠,同时提高解缠的精确性和速度。利用本论文提出的算法,根据ERS-1/2提供的SAR数据,利用InSAR技术生成了新疆喀什地区的相对数字高程图;根据ENVISAT卫星提供的ASAR数据,利用D-InSAR技术生成了反映伊朗BAM地区由2003年12月26日地震引起的地表形变的差分干涉图。更多还原

【Abstract】 Interferometry Synthetic Aperture Radar(InSAR) is an advanced space observation technique. It is the synthesis of conventional SAR techniques and interferemetry techniques that have been developed over several decades. It relies on the processing of two SAR images of the same portion of the Earth’s surface obtained either from two displaced passes of the SAR antenna at different times (repeat-pass interferometry), or from two antennas placed on the same platform and separated perpendicularly to the flight path (single-pass SAR interferometry). The 3D information of the ground and the deformation of the elevation can be extracted by according to the interference phase information of the two SAR images and the high accurate terrain information can be generated all over the world. It has the features of fine resolution mapping, working under any weather condition, imaging in daylight or at night, wide mapping range and with lots of applications. The difference InSAR(D-InSAR) is developed based on InSAR. It can monitor the tiny variation on the surface of the land and ice on the earth. It also can obtain high accurate and high credible deformation of the earth’s surface. InSAR and D-InSAR can be used to study the change of the unreachable areas before such as the glacier and the ice barrier. And also this technology can explore the Earth’s surface hazard deformation such as earthquake, volcano, coast, flood, minitary and so on. So it has been the studying hotspot in the satellite remote sensing field.The development history, application actuality, geometry imaging principle and data processing procedure of InSAR are introduced. The principles and algorithms of the key technologies including registration of SAR images, the flat earth effect removal, filtering of the interferogram and phase unwrapping are deeply discussed and analyzed. The application of the InSAR and D-InSAR in the Earth science field are studied and the actual examples are provided.The registration of the two SAR complex images used for interferometry must be performed. To obtain better measuring results, a method is proposed to improve the registration accuracy progressively for InSAR complex image. The procedure includes calculating the approximate offsets by using EnView software, pixel level’s coarse match and sub-pixel level’s precise co-registration. The primary matching algorithm based polygon structure and edge extraction enhances matching rate greatly. Considering that the presented algorithms are based on the altitude images and ignore the integration, the algorithm by using global relaxation matching on account of the phase is proposed in the last matching step to reduce the matching errors and improve matching precision. The experiment shows the nicer validity of this algorithm.The phase information of the interferogram generated after registration includes terrains information and flat earth effect information. To acquire accurate terrain information, the latter must be removed. Aimed at the difficulty to get parameters and the low precision of the common methods, a high accurate algorithm to obtain a precision result based on frequency shift and linear interpolation technology was proposed in this paper. Taking full advantage of the frequency shift method, this algorithm can improve the resolution of the frequency shift through linear interpolation. The actual data shows the feasibility and validity of the algorithm in the processing.There is multifarious noise in the interferogram with the anisotropy characteristic. The conventional methods eliminate the noise with the rectangle filtering window. Thus the phase information is confused and the fringe information maybe damaged. Aimed at the stripe’s anisotropic characteristic, a method for InSAR interferogram filtering based on the center lines of interference-stripes is proposed. It first extracts the center of interference-stripes quickly by using CNN(Cellular neural networks) algorithm for it’s parallel feature to process data. According to the center lines the contoured filtering window is obtained. The InSAR interference phase images have 2πjumping lines. So finally the sine or cosine filtering combined with the equal phase lines filtering window is performed which can remove the noise of the InSAR phase images while avoiding damaging the phase distributing. The filtering of two fact interference phase images is implemented with this method. The experiment results with good visual feature, little residue and strong image edges holding ability show that this method is efficient.There are two main types of phase unwrapping algorithms: based on path tracing and based on the least square method. The two methods hardly give attention to both aspects of precision and speed. A layering phase wrapping method based on Delaunay triangulation net minimum cost flow (MCF) algorithm is proposed. The interferogram is divided into high quality phase areas with few residues and low quality phase areas with lots of residues. Considering the irregularity of the residue distributing and the maturity of the minimum cost flow algorithm, the MCF based on Delaunay subdivision is used to unwrap the phases in the high quality components. The region growing unwrapping algorithm is used to unwrap the phases in the low quality components starting from the points in the high quality areas which has been unwrapped with MCF. The precision and the speed of the unwrapping are improved at the same time.By using the algorithms proposed in this paper and according to the SAR data from ERS-1/2, the digital elevation model of the Kashi area is generated by using InSAR technology. According the ASAR data from ENVISAT, the difference interferogram reflecting the Earth surface deformation (caused by the earthquake happened on Dec.26, 2003) image of the Bam area in Iran is generated by using D-InSAR technology.更多还原