【作者】 蔡国林；

【导师】 李永树；

【作者基本信息】 西南交通大学 ， 地图制图学与地理信息工程， 2009， 博士

【摘要】 星载合成孔径雷达干涉测量是正在发展并极具潜力的一种空间对地观测技术,具有高分辨率、全天时、全天候、大范围等优势,已在数字高程模型重建、地表形变探测、地球物理研究等方面表现出了良好的应用前景。然而,由于采用了单天线双航过模式,其获取的DEM的精度和可靠度受到四个因素的严重制约,即时间／空间失相关、大气延迟、地形因子及热噪声。为此,本文对星载合成孔径雷达干涉技术进行系统研究,重点分析这四个影响因子,探索克服这些因素的关键理论和方法,旨在提高InSAR DEM的精度和可靠度,为InSAR的相关数据处理提供新的思路和途径。具体研究内容和相关结论有：噪声类型和特点的分析是InSAR干涉图滤波的前提,论文从小波系数能量和分布特征两个角度对干涉图进行了研究。结果表明：(1)干涉相位图中只含加性噪声,不含乘性噪声；(2)高频子带中的显著系数的分布规律并不是随机的,它呈成簇分布且在同一尺度或各尺度之间都具有一定的相关性,而噪声对应的小波系数是随机分布的,不具有相关性；(3)加性噪声经小波分解,其系数服从标准正态分布。根据干涉图噪声在小波域内的标准正态分布特征,论文将小波变换和维纳滤波的优势结合,提出了一种基于小波-维纳滤波的InSAR干涉图滤波算法；又由于同一小波尺度间干涉信号具有相关性而噪声不具相关性这一特点,本文还提出了一种基于小波相位分析的干涉图滤波算法。美国Phoenix地区高信噪比干涉图和伊朗Bam地区低信噪比干涉图的两组滤波实验表明,不论是对高信噪比干涉图,还是低信噪比干涉图,这两种算法都能获得较好的效果。研究了大气延迟对雷达传播和干涉相位的影响,结果显示,气压和湿度越大,引起的雷达传播延迟量就越大,对于干涉相位而言,大气压力变化引起的相位偏差要比由水汽变化引起的相位偏差小得多。此外,根据大气延迟项的“1／f过程”特征,本文利用1／f过程的小波模型估算了研究区六幅干涉图中的噪声和大气影响值。研究了地形坡度、坡向对InSAR DEM质量的影响。结果显示：InSARDEM的精度与地形坡度存在明显的相关性。地形坡度小于10°时,DEM的偏差变化不大且精度较高,在[10°,30°]区间时,高程精度与地形坡度呈线性反比关系,大于30°时,高程精度变得不可预知；但很难发现InSAR DEM精度与地形坡向之间存在何种规律。研究了InSAR DEM精度与空间基线、时间基线之间的关系。在一定范围内,空间基线和垂直基线越长,获取的DEM的精度越高,而在空间基线长度相当的情况下,垂直基线越长,获取的DEM的精度越高。InSAR DEM的精度或多或少与干涉对的时间间隔有关,尤其在植被覆盖地区,植被生长变化对雷达回波信号的影响很大,导致了DEM精度的降低。为消弱噪声、大气延迟及空间基线对InSAR DEM精度的影响,提出了一种基于小波变换的多基线InSAR DEM融合算法。即在消除空间基线影响后,利用1／f过程的小波模型估计噪声和大气延迟对InSAR干涉相位的贡献值,并在此基础上实现小波域内的多基线InSAR DEM加权融合。美国Phoenix地区的6个InSAR DEM的融合实验表明,该算法能够有效地降低噪声、大气延迟及空间基线的影响,获得较高精度的DEM。更多还原

【Abstract】 As a developing and promising space geodetic technique, satellite synthetic aperture radar interferometry (InSAR) has the advantages of high resolution, all day, all weather, large range and so on. A number of experiments have demonstrated that InSAR is very useful in such fields as digital elevation model (DEM) reconstruction, ground deformation detection, geophysical studies, etc. However, due to the pattern of single-antenna and repeat-pass, the precision and reliability of DEM derived by InSAR would be seriously restricted by four major negative factors, i.e., temporal and geometrical decorrelation, atmospheric delay, topography as well as thermal noise. Therefore, this thesis selects these impact factors as research topics, and explores key theoretics and methods to overcome the effects of thermal noise, decorrelation, atmospheric delay and topography. The motivation of this research is to improve both accuracy and reliability in InSAR DEM and to offer new ideas and approaches for data processing of InSAR. The main research contents and relevant conclusion are as follows:In order to filter InSAR interferogram effectively in wavelet transform domain, the thesis has analyzed the types and characteristics of interferogram noises in terms of the energy and distribution of wavelet coefficients. The results are as follows:(1) there is only additive noises without multiplicative noises in interferogram. (2) The distribution regularity of remarkable coefficients in high-frequency sub-bands is not random but clustering distribution, and these remarkable coefficients have some correlation with in the same scale or among different scales, whereas the distribution regularity of the wavelet coefficients corresponding to noises is random distribution without any correlation; (3) The wavelet coefficients corresponding to additive noises obey standard normal distribution.According to the standard normal distribution of interferogram noises in wavelet domain, an algorithm of wavelet-wiener combined (WWC) filter is proposed by utilizing the merits of wavelet transform and wiener filtering. Due to the trait that the wavelet coefficients corresponding to interference signals in a same wavelet scale have some relativity, this thesis proposes an algorithm of filtering InSAR interferogram that is based on wavelet phase analysis. For validating the effects of two algorithms, two C-band interferograms, i.e., high-SNR one over Phoenix, USA and low-SNR one over Bam, Iran are selected to carry out the experiments of filter and analysis, and the results indicate that not only to the high-SNR interferogram, but also to the low-SNR interferogram, the two algorithms can both suppress noises successfully.The effects of radar propagation and interferometric phases caused by atmospheric delay are analyzed, and the results show that the bigger pressure and humidity are, the bigger retardation of radar propagation is, and for interferometric phase, the deviation caused by atmospheric pressure change is much more less than that caused by water vapor change. In addition, according to the "1/f" characteristic of atmospheric delay in frequency domain, the value of noises and atmospheric influence of six interferograms that selected for this experiment are estimated by using the wavelet model of 1/f processes.Compared to the SRTM DEM, the InSAR DEMs are evaluated to determine the impacts of terrain slope and aspect on elevation accuracy. The results show that there is an obvious relativity between InSAR DEM accuracy and terrain slope. When terrain slope is less than 10°, the deviation of DEM changes little and the accuracy is high. The DEM accuracy degrades almost linearly with increasing slope when the terrain slope is between 10°snd 30°. The accuracy of steep slopes over 30°is unacceptable. However, the rule between the InSAR DEM accuracy and the terrain slope is hardly found.InSAR DEMs are evaluated to determine the impacts of spatial baseline and time baseline on elevation accuracy. In certain extent, the longer spatial baseline and vertical baseline are, the higher accuracy is. Moreover, when the lengthes of all spatial baselines are equivalent, the longer vertical baseline is, the higher accuracy of DEM is. In terms of temporal impact, the DEM accuracy is more or less related to the time intervals of the InSAR pairs, especially for vegetation regions. Because the changes of vegetation growth seriously affect radar echo signal and lead to the accuracy reducing of DEM.To weaken the influence of noises, atmospheric delay and spatial baseline, a multi-baseline InSAR DEM fusion algorithm based on wavelet transform is proposed. After eliminating the influence of spatial baseline, this algorithm realizes the multi-baseline InSAR DEM weighted fusion on the basis of that the values of interferometric phase caused by noises and atmospheric delay are estimated by using the wavelet model of 1/f processes. Six InSAR DEMs over Phoenix, USA are selected to carry out the experiments of fusion, and the results indicate that this algorithm can effectively reduce the influence of noises, atmospheric delay and spatial baseline, and acquire DEM with higher accuracy.更多还原