【作者】 王青松；

【导师】 梁甸农；

【作者基本信息】 国防科学技术大学 ， 信息与通信工程， 2011， 博士

【摘要】 星载干涉合成孔径雷达能够全天时、全天候获取大面积高精度数字高程模型及其它增值产品,在军事侦察、国民经济建设和科学研究中有着极其广泛的应用。本文以星载InSAR数据处理方法研究为主题,以为地面数据处理系统的设计和实现提供算法支撑为目标,系统地研究了InSAR中的数据处理问题。在充分继承前人工作的基础上,针对图像配准、相位滤波与解缠、DEM重建与平差处理、SAR正射影像生成以及产品质量评估等关键环节开展研究,提出了高效高精度的处理方法,显著地改善了处理性能。论文的主要工作和创新点如下:第三章研究了干涉SAR图像配准问题。分析了各种配准测度函数的特性,确定了以实、复相关函数作为匹配度量,设计了配准灵敏度准则,实现了配准测度函数的自适应选取。进一步,提出了一种基于联合实、复相关函数的干涉SAR图像配准方法,由于综合了实、复相关函数各自的优点,使得新方法在配准准确度和稳定性上较传统方法有所提高。另外,由于相关运算可以通过快速傅立叶变换实现,新方法还具有较高的效率。第四章研究了干涉相位滤波和解缠绕问题。在分析干涉相位的统计特性和信号特征的基础上提出了非线性相位模型概念及非线性相位的自适应提取方法,利用反映地形轮廓的非线性相位模型实现了局部窗口内干涉相位的高精度逼近。然后,研究了非线性相位模型在干涉相位滤波中的应用,与文中其它经典滤波方法相比,非线性相位补偿滤波更加快速有效。最后,研究了如何利用非线性相位模型提高相位解缠性能,给出了改进枝切法和区域生长法的具体实现步骤,实测数据处理结果展示了改进方法能有效地提高相位解缠精度和速度。第五章研究了快速高精度DEM重建方法。从DEM重建原理出发,分析揭示了干涉相位与目标点三维坐标映射关系的两个基本特性。在此基础上提出了一种快速DEM重建方法,给出了快速算法的详细步骤及关键参数的取值方法。实测数据处理结果表明在重建精度损失较小的情况下,显著提高了重建速度,验证了方法的高效性和正确性。第六章研究了DEM平差方法。从DEM重建的三个方程联立求解的输入参数出发,归纳总结了影响DEM重建精度的误差源。分析了各项误差源的特性及其传递规律,确定了影响DEM重建精度的主要因素,在此基础上给出了DEM平差模型。将DEM平差分为单航带数据集平差和多航带数据集联合平差,分别研究了这两种情况下的高精度平差处理方法,由于引入了观测数据质量加权矩阵,使得平差模型参数估计的抗噪声能力增强。仿真和实测数据处理结果表明本章方法能够有效地提高DEM精度。第七章研究了干涉SAR正射影像生成方法。讨论了星载SAR定位几何原理及影响SAR定位精度的误差源,分析了各误差源与定位误差的传递关系。按照定位的解算顺序,分别提出了快速高精度前向、后向正射影像生成方法。针对目前国际上典型的四种SAR卫星——德国TerraSAR-X、意大利COSMO-SkyMed、日本ALOS-PalSAR和加拿大Radarsat-2进行了精度分析,结果表明新方法的精度完全能够满足实际需求。当用于地形校正的DEM数据与SAR影像非同源时,研究了消除二者之间相对系统误差对正射校正影响的方法。TerraSAR-X实测数据处理结果表明新方法在保持极高精度的情况下极大地提高了计算效率。第八章研究了干涉SAR系统产品质量评估。讨论了SAR图像配准、干涉相位滤波及解缠处理算法各项评估指标的特点及适用情况。阐述了DEM精度指标定义及其计算公式。研究了基于不同类型参考数据的InSAR DEM产品质量评估方法,包括点状参考数据、线状参考数据和面状参考数据,给出了方法的详细实现步骤。利用TanDEM-X的高精度DEM数据对SRTM和ASTER的DEM产品进行精度评估,实验结果与国际上公开发表文献的结论较为一致。更多还原

【Abstract】 Spaceborne interferometric synthetic aperture radar can be all-time and all-weather to obtain high-precision digital elevation models and other value-added products over large areas, which has an extremely wide range of applications in military reconnaissance, national economic construction and scientific research. This paper, on the theme of studying data processing methods for spaceborne InSAR and on the purpose of providing algorithms for the design and implementation of ground data processing systems, made a systematic study on InSAR data processing issues. Based on the work of the formers, we studied image coregistration, phase filtering and unwrapping, DEM reconstruction and adjustment, SAR image ortho-rectification, product quality assessments and so on, and then proposed a number of high-efficiency and high-precision processing approaches of which significantly improved the processing performance. The major work and innovations in this paper are as follows:The issue of interferometric SAR image coregistration is studied in chapter 3. Here we analyzed the properties of various coregistration measure functions, made the real and complex correlation functions as the matching measures, designed the criterion for coregistration sensitivity, and achieved the adaptive selection of coregistration measure functions. Further, we proposed a coregistration method for interferometric SAR image by jointing the real and multiple correlation functions. Due to the combination of the respective advantages of the real and complex correlation functions, the new method has an improved accuracy and stability for coregistration compared to the traditional ones. In addition, as the correlation calculations can be achieved through the fast Fourier transformation, the new one still has a higher efficiency.The issues of interferometric phase filtering and unwrapping are sudied in chapter 4. By analyzing the statistical and signal properties of interferometric phase, we presented the concept of nonlinear phase model and the adaptive extraction method for nonlinear phase, and achieved high-precision approximation for the interferometric phase over a local window by using the nonlinear phase model of reflecting terrain contours. Then the application of nonlinear phase model in the interferometric phase filtering is studied. While the nonlinear phase compensation filtering is more fast and effective compared to other classic filtering methods in this paper. Finally, we do research on how to improve performance of phase unwrapping by using nonlinear phase model, and provide the concrete implementation steps of the improved branch cut method and region growing method. The real data processing results show that the improved methods can increase effectively the accuracy and speed of phase unwrapping.A fast and high-precision DEM reconstruction method is presented in chapter 5. Start from DEM reconstruction theory, we analyzed and revealed the two basic characteristics of the mapping relationship between the interferometric phase and the three-dimensional coordinate of target point. Based on this, we proposed a fast DEM reconstruction method, and provided the detailed steps of fast algorithm and the solutions of key parameters. The real data processing results show that the reconstruction speed is significantly improved only with a smaller accuracy loss, thus verifies the high efficiency and correctness of the method.DEM adjustment method is studied in chapter 6. From the input parameters obtained by simultaneously solving the three equations of DEM reconstruction, we summarized the error sources of affecting DEM reconstruction accuracy. By analyzing the characteristics and transferring regularity of various error sources, we concluded the main factors of affecting DEM reconstruction accuracy, and then provided DEM adjustment model. We divided the DEM adjustment into single-orbit data set adjustment and multi-orbit data set combined adjustment, and respectively studied the high-precision adjustment approaches in both cases. Due to the quality weighting matrix is introduced into the observational data, the parameter estimation of the adjustment model has a strong antinoise ability. The processing results from the simulation and real data show that the presented method can effectively improve the DEM accuracy.The ortho-rectification method of interferometric SAR image is studied in chapter 7. We discussed the geometric theory of spaceborne SAR geolocation and the error sources of affecting SAR geolocation accuracy, and analyzed the transfer relationship between the error sources and geolocation errors. According to the solution order of geolocation, we proposed the fast and high-precision forward ortho-rectification method and backward ortho-rectification method respectively. We analyzed the accuracy of the current typical four SAR satellites: Germany TerraSAR-X, Italy COSMO-SkyMed, Japan ALOS-PalSAR and Canada Radarsat-2, and the results show that the new method can fully satisfy application requirements for the precision. When the DEM data and SAR image for terrain corrected do not have a same source, we studied the method of eliminating the effect of ortho-rectification caused by relative system error between the two. The TerraSAR-X real data processing results show that the new method not only greatly improved computational efficiency but also maintained extremely high accuracy. The quality evaluation of interferometric SAR system products is studied in chapter 8. We discussed the characteristics of the evaluation indicators and the applicable of the algorithms for SAR image coregistration, interferometric phase filtering and unwrapping. We described the definition and formula for DEM accuracy specification, studied the quality evaluation methods of InSAR DEM products with different reference data, including point reference data, line reference data and surface reference data, and provided the detailed implementation steps of these methods. The high-precision DEM data from TanDEM-X is applied to DEM products of SRTM and ASTER for accuracy evaluation, and the experimental results obtained are in good agreement with the conclusions drawn by the published sources.更多还原