多时域雷达干涉测量的优化模型及无相位解缠的参数估计方法

Multi-temporal InSAR (MTInSAR) techniques, involving the processing of multiple SAR acquisitions, provide a means to retrieve the ground height and deformation with high spatial resolution and precision. Over recent 10 years, the techniques have evolved to be an important tool for Earth observation. Currently a multitude of approaches have been proposed, which are different from each other depending on whether a unique master image is employed for the interferogram stack, whether the observation is the phase at the coherent point or the phase difference at the point-pair, and whether the parameters of interest are resolved by linear inversion or non-linear inversion methods. However there are some limitations in the available MTInSAR techniques. First, the models employed by all the methods are built up for estimating the height and deformation of the ground target simultaneously without the consideration of their sensitivity discrepancies to the spatial and temporal baselines. The joint model might result in either the loss of spatial resolution in the estimated products, which can be typically found in PSI-like techniques or the loss of estimation precision (especially the precision of height parameter),which can be typically found in short-baseline InSAR techniques. To overcome these limitations, the project dedicates to seek an improved model for the multi-temporal InSAR phases and resolve the parameters in a robust way. The main objectives include that (1) building up a MTInSAR model that has high sensitivity to the target height and deformation respectively based on different image pair combinations; (2) evaluating the ability of points and point-pairs on suppressing the effects of spatially correlated atmospheric delay under the framework of MTInSAR; (3) designing a solver based on L1 norm criteria to estimate the parameters without phase unwrapping. The proposed model and parameter estimator has the ability to improve the precision and the reliability of the MTInSAR results which are valuable for the assessment of Geo-hazards.

现有的多时域雷达干涉测量技术（MTInSAR）均按同时解算高程和形变参数的思路建模。由于没有考虑待求参数及相干点密度对基线（时间，空间）的敏感性差异，所构建的模型要么造成参数估计的空间分辨率较低（如永久散射体技术）要么造成参数（特别是高程参数）估计精度的降低（如短基线集技术）。在参数解算上现有方法均需考虑相位模糊度的影响使得解的成功率难以保证。本项目旨在探索MTInSAR技术的优化模型及稳健的参数解算方法。研究内容包括：（1）建立基于不同基线组合分别求解相干点高程及形变的优化模型；（2）在MTInSAR框架下评估点及弧段对空间相关的大气分量的抑制能力及对参数解算精度的影响；（3）建立基于L1范数准则的无需相位解缠的参数解算方法。本项目意义在于所提出的模型和参数解算方法可以提高MTInSAR形变探测空间分辨率、精度和可靠性，所获取的地表时空形变信息将为地质灾害的监测和防治提供有力的数据支撑。

星载雷达干涉测量技术（Satellite SAR Interferometry， InSAR）作为一种新兴的空间对地观测技术，在对地质灾害引起的地表沉降监测方面极具潜力。然而干涉相位作为InSAR技术的基本观测量，受到多种因素的干扰，在实际应用中测量精度及可靠性难以保证。研究发现，通过分析同一区域不同时刻获取的一系列雷达影像可获得更可靠且稳定的形变监测成果。这种分析方法称之为时序干涉测量技术，但现有的时序技术无论以单一主影像干涉序列为观测量还是以多主影像干涉序列为观测量，均在同一时空基线组合的框架下同时求解DEM误差和形变参数。由此带来的问题要么是由于时间空间失去相干而造成相干点密度极大降低，要么是所获得的DEM误差度不高。本课题围绕这些问题开展了四个方面的研究，即建立和评估基于不同时空基线组合的参数解算模型；在MTInSAR框架下评估点和点对对空间相关的大气分量的抑制能力；建立基于L1范数准则的具有抗差性的参数解算方法及评估所建立的优化模型及参数解算方案的性能。通过大量的模拟实验和实际数据，我们系统分析了影响参数解算的时空基线组合，证实形变和高程参数对时空基线的响应是不同的。现有的时序模型由于均在相同的观测下同时解算这两类参数，无法高精度地实现参数解算，因此我们提出了一种用于高程参数估计的时序模型并成功应用于滑坡监测。针对不同尺度的大气延迟误差，我们分析了点与点对（或弧段）对其抑制能力，并明确了各自的优缺点。在参数解算方面，基于现有的观测模型，我们实现了两种抗差估计方法，并比较了各自的优劣及与已有解算方法的差异。该估计方法已整合到了自主开发的TCPInSAR处理平台并利用加州地区近20年的雷达数据及GPS数据进行了测试和验证，效果令人满意。基于项目成果，本课题组共发表论文11篇其中SCI论文10篇，完成软件著作权登记1项，申请算法专利1项。主要存在的问题是经费使用相对于研究进度较为滞后，特别是在劳务费方面上，今后将会高度重视这类问题。