【作者】 郭交；

【导师】 保铮；

【作者基本信息】 西安电子科技大学 ， 信号与信息处理， 2011， 博士

【摘要】 获取地球表面的高精度数字高程模型(Digital Elevation Model, DEM),在军事应用以及国民经济建设方面都具有重要的意义。干涉合成孔径雷达(Interfero-metric Synthetic Aperture Radar, InSAR)在生成数字高程模型方面具有巨大的潜力和广阔的应用前景,以分布式卫星为平台,能够快速高效、大范围地获取地球表面的DEM,完成全球测绘,具有其它测量手段无法比拟的优势。本论文以分布式卫星InSAR获取地球表面数字高程模型为主线,重点研究了分布式卫星InSAR信号处理的几个关键技术。全文总体上分为两个部分：第一部分主要研究了单基线分布式卫星InSAR技术,对于普通地形,利用传统的单基线InSAR能够有效地获取这些区域的DEM。然而,对于复杂地形(例如坡度较大的局部区域、山谷和悬崖等),由于存在严重的相位欠采样和高程层叠效应,使这些区域成为单基线InSAR测绘的“盲区”,而多基线InSAR技术是解决此问题的一项关键技术,因此论文第二部分主要针对多基线分布式卫星InSAR技术展开了研究。本文的主要工作概括如下：1.提出了一种基于相同方位角的方位向预滤波方法。InSAR处理要求用于干涉的SAR图像之间具有很高的相干性,相干性直接决定DEM的高程测量精度。针对空间基线(包括垂直航向基线和沿航向基线)引入的去相干因素,通过对回波数据进行距离向和方位向预滤波处理,能够有效提高SAR图像之间的相干性。我们在对InSAR成像的几何关系和信号模型详细分析的基础上,揭示了雷达回波之间具有相干性的本质,针对方位向预滤波处理,提出在方位向只有具有相同方位角的空间采样位置接收到的雷达回波才具有相干性,能够用于后续的相干干涉处理,并在此基础上提出了一种基于相同方位角的方位向预滤波方法,通过计算机仿真验证了此方法的有效性。2.提出了基于粗DEM的SAR图像配准—干涉相位滤波算法。SAR图像配准和干涉相位滤波是InSAR处理中的两大步骤。一方面,InSAR处理要求SAR图像之间的配准精度很高,但实际处理中由于逐点精配准要求的计算量巨大,传统的图像配准算法都是生成少量的控制点,利用一定准则求取控制点的二维偏移量,再通过一定的模型拟合获得整幅二维图像中每个像素的偏移量。但是对于高程起伏剧烈的地形,或者大幅图像进行配准时,通过简单的多项式拟合求取偏移量会引入较大的误差。另一方面,干涉相位滤波算法总是假定滤波窗口内的样本点满足独立同分布的假设,然而实际中受地形起伏的影响,滤波窗口内的样本点不可能严格满足独立同分布的假设,导致滤波结果偏离真实值。针对上述问题,我们的思路是充分利用已有的粗DEM来提高SAR图像配准和干涉相位滤波的性能,并大大减少运算量。在SAR图像配准中,首先利用粗DEM完成对SAR图像中的每一像素的定位,然后在SAR图像中逐点求取每个像素的二维配准偏移量,从而完成整幅SAR图像的配准。在干涉相位滤波中,为了使滤波窗口内的样本数尽量满足独立同分布,我们利用已知的粗DEM对干涉相位图中的地形信息进行统一补偿,以获取更多的独立同分布样本。3.研究了联合单航过和多航过InSAR数据的处理方法。对于坡度平缓的地形,利用单基线InSAR生成的DEM,其精度已经足够,但对于高复杂地形的DEM测绘,则需要利用多基线InSAR技术。在实际中,利用现有的单基线InSAR系统对同一地区进行多次航过重复观测,是一种现实可行的获取多基线InSAR数据的途径。针对以单基线InSAR多次航过获取的多基线干涉数据,我们研究了多次航过之间的干涉基线估计方法和基于干涉相位图的图像配准方法。在干涉基线估计方法中,首先对多次航过的数据分别进行精确的SAR成像、图像配准和干涉相位滤波处理,然后联合两次单航过的相位图对局部高相干区域进行稳健的多基线相位解缠,获得对应目标点的高程信息,最后结合多次航过间的干涉图对多次航过的干涉基线进行解模糊和估计。在干涉相位图的配准方法中,对单航过生成对应的干涉相位图进行配准,而不是直接配准SAR图像,从而避免了多航过SAR图像之间相干性较低导致配准难度大的问题。4.提出了一种利用多基线InSAR技术恢复高程层叠区域高程信息的方法。首先研究了在传统窄带阵模型假设条件下利用多基线InSAR解决高程层叠问题的方法,并指出随着SAR图像分辨率的提高和基线长度的增加,窄带阵模型并不完全适用于实际的星载InSAR系统,在此基础上提出了更加符合实际情况的宽带阵模型,并针对此模型提出了两种解决方案,即地形搜索的方法和联合距离像素处理方法,减小甚至消除了层叠的多个地面目标回波包络未完全配准带来的影响,提高了高程估计的稳健性。更多还原

【Abstract】 It is of great importance to acquire digital elevation model (DEM) of the Earth’s surface with high accuracy in the fields of military application as well as national economic construction. Interferometric synthetic aperture radar (InSAR) has tremendous potential and broad prospects in generating highly accurate DEM. Distributed satellite InSAR is capable of acquiring large-scale DEM effectively and accomplishing global DEM generation, possessing fabulous advantages compared with other topography measuring methods.The dissertation, with DEM generation by distributed satellite InSAR as the main line, focuses on several key techniques of signal processing. The whole dissertation is composed of the following two parts:In the first part, the single-baseline distributed satellite InSAR (i.e., single pass of two formation-flying satellites) technique is studied. For general topographies, the conventional single-baseline InSAR system with two antenna phase centers has been proved to have the capability to provide the DEMs with high accuracy. However, for complicated topographies containing highly sloping regions or discontinuous surfaces (such as man-made buildings, canyons and steep mountains that exist numerously on the Earth), the performance decreases severely due to the serious undersampling phases and layover phenomena. In this case, the multibaseline InSAR is an effective technique to overcome the drawbacks. Therefore, the second part investigates the multibaseline distributed satellite InSAR (including multi-pass of two formation-flying satellites and single-pass of multiple formation-flying satellites).The main work of the dissertation is summarized as follows:1. In Chapter 2, an improved method for InSAR azimuth prefiltering is proposed based on the principle of coherence. InSAR data processing requires high coherence between the interferometric SAR images, which directly determines the accuracy of the DEM. For the decorrelation induced by the spatial baseline (including the along-track and cross-track baselines), the azimuth and range prefiltering is performed to increase the coherence effectively. Based on the theoretic analysis of the InSAR imaging geometry and signal models, the principle of coherence is revealed, and then we propose that only the radar echoes received by the space sampling positions with the same azimuth angles are coherent, i.e., available for interferometric processing. An improved method for InSAR azimuth prefiltering is proposed according to the same azimuth angle extension. The computer simulation is also carried out to prove that the proposed algorithm has the ability to improve the coherence between interferometric SAR images.2. Chapter 3 presents SAR image coregistration and interferometric phase filtering algorithms based on coarse DEM. On the one hand, InSAR data processing requires high accuracy for the SAR image coregistration. Due to the heavy computation burden, the conventional coregistration algorithms compute the two-dimensional offsets by polynomial fitting of a few control points. However, for complicated topographies or large-scale scenes, the offsets obtained by simply polynomial fitting results in unexpected errors. On the other hand, the phase filtering algorithms always make the assumption that the samples in the local window obey the independent and identically distributed (i.i.d) conditions. But in fact, the obtained samples in the filtering window do not always satisfy the i.i.d conditions due to the terrain changes over the window, thus degrading the phase filtering performance. In order to deal with the above problems, we propose the method that makes full use of the available coarse DEM to assist the operations of SAR image coregistration and interferometric phase filtering with the purposes of increasing the performance and decreasing the computation burden. The proposed method firstly computes the two dimensional shift amounts of each pixel in SAR images according to the coarse DEM and InSAR system parameters. For phase filtering, in order to obtain more i.i.d. samples, we use the interferogram generated by the coarse DEM to compensate the terrain phases of the whole scene.3. InSAR data processing in the combination with the single-pass and multi-pass data is investigated in Chapter 4. For the areas with moderately sloped topographies, the conventional single-baseline InSAR system has the ability to generate the DEMs with high accuracy. However, in order to obtain DEMs for hilly and mountainous areas, the multibaseline InSAR is necessary. There exist various ways to acquire multibaseline InSAR data. In practice, it is a feasible way to obtain multibaseline interferometric data by employing repeat passes of a single-baseline InSAR system that exists already. In this case, the methods for estimating the effective baseline length and coregistering interferograms between multiple pass SAR images are proposed. For baseline estimation, the proposed method unwraps local region phases to acquire the heights of multiple ground scattering units by combining multi-pass interferometric data, and then makes full use of multiple targets’heights and the multi-pass interferogram to estimate the effective baseline length. For interferogram coregistration, an innovative strategy is proposed based on the phase gradient, which avoids the difficulties induced by the serious temporal and spatial baseline decorrelation for the coregistration operation of multiple pass SAR images.4. Chapter 5 deals with the problem of retrieving heights of layovered terrains using multibaseline InSAR. On the basis of studying conventional methods resolving the layover problem, the drawbacks under the narrow band assumption are revealed, especially for the case of high resolutions and long baselines. In order to complete the envelope alignment, the narrowband array echo model is substituted by the wideband array signal model. Two strategies to solve the wideband array model are proposed, with one called the aligning method and the other the joint range cell processing method. Both of the two strategies have the ability to eliminate or mitigate the effects of envelope misalignment and to increase the robustness of height estimation greatly.更多还原