【作者】 查显杰；

【导师】 傅容珊；

【作者基本信息】 中国科学技术大学 ， 固体地球物理学， 2007， 博士

【摘要】 InSAR(Interferometric Synthetic Aperture Radar)是一种新型空间对地观测技术，它具有高空间分辨率、高精度和面矩阵测量等优点，受到国际地学界的广泛关注。然而SAR采用侧视成像工作方式，其形变干涉图只含有地表形变在雷达视线向的信息，因此关于InSAR形变测量结果的解释和应用等问题还有待于进一步研究。本文基于InSAR成像理论，提出了利用卫星轨道数据和形变模型模拟地表形变场干涉图的方法。在此基础上，设计了三个理想位移模型并模拟得到了相应的三个不同方向的形变SAR干涉图。对比模拟干涉图中得到的形变位移量与理想形变模型得出：差分干涉图对不同方位形变的敏感性存在明显差异；对竖直向形变最敏感，距离向次之，对方位向形变最不敏感。考虑到SAR干涉图与GPS、精确水准测量等常规大地测量结果之间缺乏直观可比性，同时干涉图不便直接用于其它相关研究，本文提出了直接从三个不同视角形变干涉图中获取三维形变位移场的方法。为了检验该方法的正确性，本文首先采用数值模拟方法模拟了三个不同视角的同一形变场的SAR干涉图，然后利用相位场解缠算法得到了形变场在三个不同视向的位移量，再通过矢量分解获取了形变位移场的三分量。结果显示：反演得到的地表形变位移三分量基本能再现形变场模型，且三个分量的误差均小于15％。理论和模型计算表明利用形变干涉图直接获取形变位移三分量是可行的。基于我们提出的直接从三个不同视角的形变干涉图中计算三维形变位移场的方法，本文获取了2003年伊朗Bam地震的三维同震形变位移场。利用欧空局提供的7景Bam地震的Envisat卫星SAR数据，采用差分干涉算法处理获取了两幅降轨同震形变干涉图和一幅升轨震后形变干涉图。两幅降轨同震形变干涉图中干涉条纹的大小和形状都比较类似，说明它们反映的几乎是同一形变位移场。由于生成这两幅降轨干涉图的主像相同，而从像视角略有差异，因此可以认为这两幅形变干涉图是视角不同的同一地表形变干涉图。基于以上分析，本文通过配准干涉像对的幅度影像，获取了形变位移场在方位向的分量，然后将其分解得到了地震形变水平位移场，结合从解缠相位场中得到的雷达视线向位移，获取了形变位移场的竖直向分量。水平位移场显示Bam地震主断层为隐伏右旋走滑断层，这与整个区域的大地构造背景相吻合。同时根据水平位移矢量的方向差异，我们勾绘出了该隐伏的地震断层在地表的走向和格局，显示Bam地震对应的隐伏断层呈“Y”字形分布，与Nakamura 2005给出的结果基本吻合。结果证实了从两个不同视角的干涉像对和形变干涉图中获取形变位移三分量的可行性。由于干涉像对的基线和时间间隔过长以及大气扰动等干扰因素的影响，实际的InSAR干涉图常含有大量噪声。大多数常用去噪算法直接对干涉图操作，但是在很多情况下不但去噪效果不好，而且会导致有用信息的损失或歪曲。利用小波多分辨分析能有效刻画噪声等非平稳信号并使噪声趋于白化的特点，本文提出了以小波包变换为基本算法，利用Wiener滤波器在小波域对复干涉图实部和虚部分别去噪的新方案。通过选用不同的小波基函数和加噪模拟干涉图实验发现：Daubechies小波和双正交样条小波干涉图去噪效果相当，多级小波包分解去噪比一级小波包分解效果更好，但会损失相对较多的边缘信息。为了进一步检验本文方案的有效性，我们采用本文方案、Wiener滤波方法和小波包软阈值法对两幅含噪水平不同的实际干涉图进行了实验。结果表明在小波域采用Wiener滤波方法既能有效消去干涉图中噪声，又能最大限度的保留干涉图的边缘信息，去噪效果明显优于后两者。为了解决人口稀疏和地震观测能力不足地区的地震定位问题，我们尝试了利用InSAR准确测定地震震中位置的方法。目前，基于InSAR技术定位地震震中还很不成熟，尚在发展之中。对于强烈地震而言，震中地区地表破裂区域较大，破坏严重，地面形变格局复杂，故在震中区域的InSAR干涉条纹不清晰，很难准确确定震中位置。而对于小地震，地表形变往往太小，或是InSAR数据获取时间间隔过大，采用差分干涉方法不能有效地监测。中等强度的浅源地震对应的地面形变较为平缓，其SAR数据相对于强烈地震和小地震来说相干性较好，同震形变场干涉图噪声小、条纹比较清晰。通常距离震中越远形变越小，干涉图的条纹率越低，因此可以根据形变场干涉条纹的分布来确定震中位置。本文利用1999年3月15日新疆库车M_W5.4地震的三景SAR数据，对基于InSAR技术定位地震震中进行了探索性研究。首先利用SAR数据生成了库车地震同震形变场差分干涉图，然后采用精确的地理编码方法对差分干涉图进行地理编码，从而得到具有大地坐标的形变场干涉图。不过，库车地震的三景SAR数据组成的像对时间基线距比较大，干涉图中大部分区域失相干严重，本文只能根据同震形变场差分干涉图中仅存的干涉环分布，初步确定库车地震震中位置为(82°36′E，41°53′N)。这一结果与Harvard大学给出的震中位置相距20公里左右，与新疆地震局、USGS给出的震中位置仅相差1.5公里左右的距离，并且本文的定位结果位于该地震等震线的最内层之中，表明InSAR方法定位结果基本上是可信的。由于本文所采用的SAR数据受到限制，没有得到高质量的地震同震形变干涉图，使InSAR技术定位地震震中的优点没能得到充分体现。我们相信在不久的将来，随着高质量SAR数据的获取，InSAR技术将在地震的精确定位方面发挥重要作用。更多还原

【Abstract】 InSAR （Interferometric Synthetic Aperture Radar） is a novel space observation technique from satellite which enjoys the merits of high spatial resolution, high precision and area matrix surveying, therefore absorbs many geo-scientists’ attention. However, the deformation interferograms only contain the surface deformation information in the radar’s line of sight because of the side-looking work mode of SAR. Many problems including the explanation and application of the InSAR interferograms need further research.This paper firstly proposed a simulation method for the interference pattern of surface deformation field using the satellite orbit data and deformation models, based on the image principle of InSAR. Then we constituted three ideal displacement models and obtained the SAR deformation interferograms of three different orientations. Comparing the derived displacements from interferograms with deformation models, we can draw a conclusion that the sensitive of interference pattern to deformation in different orientation is obviously different and the sensitivity value from large to small is vertical, range and azimuth. Considering the interference pattern can not be directly compared with the results derived from GPS, spirit leveling measurement and other general geodetic method, and can not be directly applied to relative researches, a new method of directly acquiring three displacement components from surface deformation interferograms with three different look angles has been proposed. In order to test the efficiency of the method, we modeled SAR interferograms of the same deformation with three different look angles, unwrapped them with phase unwrapping algorithm, thus obtaining the displacements of three different look angles, and finally acquired 3D displacements through the decomposing method of displacement vectors. The results show that the inversed three components of the surface deformation displacements are basically consistent with the models and the errors are less than 15%. Theory and model calculations demonstrated that it is feasible to directly obtain three components of the surface deformation displacement from interferograms.Based on the above-mentioned calculation method, we acquired the 3D coseismic deformation displacement field of the 2003 Bam earthquake, Iran. Using 7 scene ASAR SAR images of Envisat satellite from European Space Agency, we got two descending interferograms and one ascending interferogram through processing by interferometric method. In two descending interferograms, the similarity of the size and shape of interferometric fringes show that they reflect the nearly same deformation fields. Because the main images in interferometric pairs are identical and the slave images are different in look angle, the interferograms derived from those pairs can be considered as the same deformation field’s observation with different look angle. Based on the above analysis, we obtained the 3D displacements through the following three steps. Firstly, we retrieved the azimuthal displacements by registered the intensity images in interferometric image pairs. Then by decomposing the azimuthal displacements, we attained the horizontal displacement field. Finally we obtained the vertical displacements from the horizontal and the displacements in the line of radar sight derived from unwrapped phase. The horizontal displacement field suggests that Bam fault is of a right-lateral and strike-slip-type blind fault, which is consistent with the geotectonic background of the whole area. According to the difference of the horizontal displacement vector direction, we described the strike and pattern of the surface rupture. The rupture is "Y" shaped, which is basically consistent with the results given by Nakamura 2005. The research results proved that 3D displacements could be obtained from two interferometric pairs with different look angle and deformation interferograms.The real interferograms are affected by noise, introducing by too long baseline or temporal interval of the interferometric pair and the interfering of atmosphere. Many prevalent noise removal algorithms directly process the interferograms, which can not efficiently suppress noise and may make the useful information loss or distorted. The wavelet multi-solution analysis can effectively represent the no-stationary signal such as noise and make the noise whiten. This paper proposed a new noise reduction scheme, adopting wavelet packet transform as base algorithm and denoising the real and imaginary part of complex interferogram respectively by Wiener filter in wavelet domain. By choosing the different wavelet base function, the filtering results of the simulation interferograms show that the noise reduction effects of interferogram are similar by Daubechies wavelet and Biorthogonal wavelet. In addition, multi-level wavelet packet decomposition precedes one-level wavelet packet decomposition, but the former can make the more useful information loss. To further test the validity of the scheme proposed in the paper, we processed two real interferograms with different noise level using our scheme, Wiener filter and wavelet packet soft thresholding method. The results show that our scheme, which can not only effectively remove noise in interferogram but also preserve the fringe of phase image to the greatest possible extent, clearly outperforms Wiener filtering method and wavelet packet soft thresholding method.To locate the epicenter in underpopulated area with earthquake observation deficiency, we attempt to investigate the precise epicenter location method using InSAR. Presently, it is not mature and under developing to locate the epicenter of earthquake based on InSAR technique. For large earthquakes, the surface ruptures in epicenter are considerably strong and destroy heavily. Therefore the surface deformation pattern may be complex and make the fringes in epicenter area illegible and make it difficult to accurately determinate the epicenter location. For small earthquake, the deformation is too small or the interval of the interferometric pairs is too long, so the surface deformation may not be detected by interferometric method. Fortunately, the deformation of the moderate-sized shallow earthquake is moderate in the epicenter area and adapt to the observation of InSAR. Moreover, the forming fringes are always very clear and less noise. Generally, the farther to the epicenter, the smaller the deformation, the lower the fringe rates of the interferogram, so the precision epicenter location can be determined according to the distribution of interferometric fringe of deformation field. In this paper, we studied the earthquake epicenter locating method based on InSAR technique using three scene SAR images of the 15^th March 1999 Mw=5.4 Kuche earthquake, Xinjiang province, China. We firstly generated the coseismic deformation field interferogram of the Kuche earthquake from SAR data. Then we coded the interferogram using the accurate geo-coding method and acquired the coordinates of deformation field interferogram. However, the intervals between three SAR images of Kuche earthquake are too big, so it is seriously decorrelation in most regions. According to the distribution of the very limited interferogram fringe, we preliminarily determinated the epicenter location of Kucheearthquake at （41°55’N,82°35 E）. Comparing the epicenter locations given by us, Harvard,USGS, Xinjiang earthquake bureau, we found that the locations of us, USGS, and Xinjiang earthquake bureau are very approximate and the three locations are less than 1.5km apart. The Harvard’s location result deviates badly and seems wrong, whose site apart the others is about 20km. In addition, the epicenter determinated by us locates the most inner isoseismal curves of this earthquake. Those show that the InSAR epicenter location method is credible. Because the available SAR data is of low-level quality and limited, we could not obtain ideal coseismic deformation interferogram, thus the advantages of InSAR epicenter location method are still not adequately revealed. We believe that with high-level quality SAR data available, the InSAR technique will play an important role in precision epicenter location in the near future.更多还原