【作者】 张红敏；

【导师】 徐青；

【作者基本信息】 解放军信息工程大学 ， 摄影测量与遥感， 2013， 博士

【摘要】 近年来我国社会、经济平稳快速发展，对基础地理信息的现势性和精度需求不断提高。SAR（Synthetic Aperture Radar，合成孔径雷达）图像定位与三维信息提取是获取基础地理信息的重要技术途径之一，具有全天时、全天候、高精度等突出优势。但在SAR图像定位与三维信息提取中仍存在诸多技术难题。为了提高我国SAR图像的定位精度和三维信息提取能力，本文重点对立体SAR图像定位、单幅SAR图像无控制定位、多基线/MIMO（MultipleInput Multiple Output，多输入多输出）InSAR（Interferometric Synthetic Aperture Radar，干涉合成孔径雷达）图像定位与三维信息提取、SAR图像匹配等问题进行研究，采用高分辨率机载SAR图像进行了大量实验。论文完成的主要工作和创新点有：1．全面梳理了SAR图像定位与三维信息提取技术的研究现状，剖析了制约SAR图像三维信息提取实用化的主要问题，结合对SAR图像几何特性与辐射特性的分析，辨析了相干斑与噪声、图像分辨率与像元采样间隔等易混淆的概念，论述了SAR图像几何构像模型，介绍了立体SAR图像定位、单幅SAR图像定位、干涉SAR图像定位与三维信息提取的基本原理。2．针对高分辨率机载斜侧视立体SAR图像，设计了标定近距延迟和多普勒频移参数的R-D（Range-Doppler，距离-多普勒）模型高精度定位方案，克服了定向参数求解难题，实现了1米量级误差的机载SAR图像立体定位。3．提出了DEM（Digital Elevation Model，数字高程模型）辅助的单幅SAR图像无控制斜侧视定位方案。该方案通过DEM仿真SAR图像，并将仿真SAR图像与实际斜侧视SAR图像进行匹配来提取控制点，采用斜侧视R-D模型计算出定向参数后，在DEM的支持下完成单幅SAR图像的定位。针对仿真SAR图像与实际SAR图像的匹配问题，研究了分别以仿真SAR图像、实际SAR图像作为匹配基准的两种匹配策略，在实验分析基础上，得出了以仿真SAR图像作为匹配基准效果较好的结论。采用高分辨率机载SAR图像，利用SRTM（Shuttle Radar Topography Mission，航天飞机雷达地形测绘任务）DEM或地球模型方程进行了一系列单幅SAR图像定位实验，统计分析了各种单像SAR定位方法的定位误差，验证了所提出方法的有效性。4．提出了差分滤波MIMO InSAR图像定位与三维信息提取方案。利用DEM，完成了直接法和间接法干涉图仿真，并采用仿真的不同基线干涉图进行了单基线InSAR定位与三维信息提取实验，验证了短基线InSAR精度低、可解性强，长基线InSAR精度高、可解性差的特点；设计了差分滤波多基线InSAR定位与三维信息提取方案和中国余数定理多基线InSAR定位与三维信息提取方案，采用仿真数据进行了实验，验证了多基线InSAR高精度、高可解性的优势；提出了差分滤波MIMO InSAR定位与三维信息提取方案，由最低频干涉图开始进行滤波和相位解缠，并依次对较高频干涉图进行差分、滤波和相位解缠处理，从而完成高精度的高程信息反演。采用仿真的MIMO InSAR干涉图进行了干涉处理实验，验证了方案的可行性。5．针对SAR图像间相对几何畸变和辐射畸变较大的问题，提出了利用POS（Positionand Orientation System，定位定姿系统）和SRTM DEM数据进行点位预测的机载SAR图像多视双向自动匹配方法。该方法利用POS和SRTM DEM数据进行点位预测，并采用强度影像或其线特征增强影像，进行机载SAR图像的多视双向自动匹配。采用多种不同分辨率的机载SAR图像进行匹配实验，得到了较好的匹配效果。6．集成了立体SAR图像定位、单幅SAR图像定位、干涉SAR图像定位与三维信息提取、SAR图像匹配等研究成果，构建了SAR图像定位与三维信息提取试验软件系统，采用中国测绘科学研究院获取的机载SAR图像和仿真SAR图像进行了大量定位、三维信息提取和匹配实验，验证了论文研究成果的正确性和有效性。在SAR图像高精度定位与三维信息提取方面取得的研究成果，将提高我国SAR图像的定位精度，促进我国SAR图像三维信息提取技术的发展和SAR技术在地形测绘中的应用。更多还原

【Abstract】 The economy and society in China develop steadily and fleetly in recent years. It demandsto improve the up-to-data efficiency and the precision of fundamental geography information.Positioning and three dimensional (3D) information extraction from Synthetic Aperture Radar(SAR) images, is an important approaches for deriving fundamental geography information. Ithas the advantages of all time, all weather, high efficiency and high resolution. However, thereare many technology difficulties in positioning and3D information extraction from SAR images.In order to improve the capability of positioning and3D information extraction, positioningmethods with stereo SAR images as well as single SAR image without Ground Control Point(GCP) were studied. Positioning and3D information extraction from multi-baseline or MultipleInput Multiple Output (MIMO) Interferometric SAR (InSAR) images, as well as matching ofSAR images were researched. Experiments using high resolution airborne SAR images weredone. The main work and innovations of this thesis are listed as follows:1. Comprehensive review of the present situation of positioning and3D informationextraction from SAR images was done. The main problems that restrict the utility of3Dinformation extraction from SAR images were analyzed. With the geometry and radiationcharacter analysis of SAR images, the confusing concepts of speckle and noise, image resolutionand pixel spacing were analyzed. Geometrical imaging models of SAR images, positioning withstereo SAR images, positioning with single SAR image, and the fundamental principles ofInSAR images for positioning and3D information extraction were systematically discussed.2. An accurate positioning scheme based on Range-Doppler (R-D) model was designed forhigh resolution airborne squint stereo SAR images. It calibrates the close range and the Dopplerfrequency as overcomes the difficulty in solving the orientation parameters. Airborne SARimages positioning with1meter level error was realized.3. A positioning scheme for single squint SAR image based on Digital Elevation Model(DEM) without GCP was proposed. It simulates SAR image with DEM, and extracts GCPs bymatching the simulated SAR image and the actual squint SAR image, and then, the orientationparameters with squint R-D model can be calculated to complete single SAR image positioningsupported by DEM. For the matching of simulated SAR image and actual SAR image, two kindsof matching strategies that respectively take simulated SAR image or actual SAR image asreference were discussed. The conclusion is that the matching with simulated SAR image asreference is better. A series of positioning experiments with single SAR image were carried outusing high resolution airborne SAR image, Shuttle Radar Topography Mission (SRTM) DEM or earth model equations. The errors of positioning with each scheme were statistically analyzed, asverified the proposed method.4. A scheme with differential filter for positioning and3D information extraction fromMIMO InSAR images was proposed. The interferograms were simulated with direct algorithmand indirect algorithm. The experiments of positioning and3D information extraction fromsingle baseline InSAR were carried out with simulated different baseline interferograms. Itverified the fact that the shorter baseline InSAR has lower accuracy and strong solvability, whilethe longer baseline InSAR has the opposite feature. Multi-baseline InSAR can utilize variousbaseline data. The scheme with differential filtering and the one with Chinese RemainderTheorem (CRT) for multi-baseline InSAR were designed. The advantages of higher precisionand solvability in multi-baseline InSAR were verified through the experiments with simulateddata. The scheme with differential filter for positioning and3D information extraction fromMIMO InSAR images was proposed. The interferogram with lowest frequency was processedfirstly by filtering and phase unwrapping, and then the interferogram with higher frequency wasprocessed by difference, filtering and phase unwrapping one by one, so as to finish accurateheight deriving. MIMO InSAR interferograms simulation and processing experiments werecarried out, as validated the proposed scheme.5. An automatic matching for SAR images with multi-look and bilaterial sequence byprediction of point’s position with Position and Orientation System (POS) and SRTM DEM datawas proposed. This method predicts point position using POS and SRTM DEM data, enhancesimages by extracting line features, and proceeds with multi-look and bilaterial automaticmatching. It can decrease the effect of the larger relative geometric and radiation distortion. Itcan get better matching results, as was verified by varied different resolution images.6. The research achievements were integrated, as including stereo SAR images positioning,single SAR image positioning, InSAR images positioning and3D information extraction, SARimage matching and so on. And the software for SAR image positioning and3D informationextraction was built. Experiments of positioning, extraction of3D information and matching ofSAR images were done with airborne SAR images acquired by Chinese academy of surveyingand mapping and with simulated SAR images, as validated the research achievements.The research achievements in accurate positioning and3D information extraction from SARimages will improve the positioning accuracy of SAR images in our country. It can promote thedevelopment of3D information extraction from SAR images and the application of SAR interrain surveying and mapping.更多还原