摘要:
合成孔径雷达干涉测量（interferometric synthetic aperture radar，InSAR）技术可获取广域高精度地表形变量，被广泛应用于地质灾害监测等领域，但是其监测性能取决于散射体在不同观测时间之间散射特性的相关程度。相位连接（phase linking，PL）方法利用多时相干涉相位分析恢复系统相位序列，是解决时序InSAR（time-series InSAR，TSInSAR）低相干地表监测难题的关键所在。近十几年来，学者们提出了一系列行之有效的实现方法，其算法差异可归纳于所采用的定权策略不一致。从研究动机、统计基础、方法进展和结果分析等多个方面介绍了现有PL方法在TSInSAR地表形变监测等领域的研究进展，并对未来发展趋势进行了讨论。首先，从干涉相位分量和失相干源出发，分析永久散射体与分布式散射体相位一致性差异及其原因，并指出相位连接研究的必要性；其次，在介绍复协方差矩阵及其统计分布的基础上，对现有PL方法进行归类和对比，并指出极大似然估计器的优势及局限性和相位优化理论精度上限；然后，结合蒙特卡洛模拟实验和真实数据实验，定性和定量分析现有PL方法在干涉相位优化、形变测量精度和算法计算效率等方面的差异；最后，总结了PL方法在TSInSAR形变监测应用的局限性，并讨论其未来发展方向。
Abstract:
Objectives:Interferometry synthetic aperture radar (InSAR) can acquire wide-area high-precision surface morphology, which is widely used in the field of geohazard monitoring, etc. However, its monitoring performance strongly depends on the correlation degree of scattering characteristics of scatterers between different imaging moments. The phase linking (PL) method utilizes multi-reference interferometric phase analysis to recover the systematic phase series, which is the key to solving the problem of time-series InSAR (TSInSAR) coherence-low surface monitoring. In recent decades, scholars have proposed a series of effective implementation methods, whose differences can be attributed to the inconsistency of the adopted weighting strategy. Methods:We introduce the research progress of the existing PL methods in the TSInSAR surface deformation monitoring and other fields from various aspects, including research motivation, statistical basis, method progress and result analysis, and then discuss the future development trend. First, from the perspective of the interferometric phase components and decorrelation sources, we analyze the differences in phase consistency between both persistent and distributed scatterers and their reasons, and point out the necessity of PL research. Second, based on the introduction of complex covariance matrices and their statistical distributions, we categorize and compare the existing PL methods and point out the advantages and limitations of the great likelihood estimator and its accuracy upper limit for phase optimization theory. Results:The simulated experiment uses the Monte Carlo simulation to generate the multilloking time-series covariance matrix under both exponential and seasonal temporal decorrelation models, and quantificationally compare five classical PL methods in terms of interferometric phase optimization, deformation measurement accuracy and algorithm computational efficiency. In the real experiment, the Daguangbao landslide triggered by the 5·12 Wenchuan Mw 7.9 earthquake and the 56 scenes of Senetinel- 1 synthetic aperture radar (SAR) images in VV polarization are selected as the experimetal data and the study area, respectively. In addtion to the persistent scatterer interferometry technology, three PL methods are qualitatively compared and analyzed in the three aspects, similar to the simulated experiment. Conclusions:The existing PL methods are summarized and their limitations in TSInSAR deformation monitoring, data compression and SAR tomography are pointed out. With the development of spaceborne SAR platforms towards multi-band, multi-polarization and high spatiotemporal resolution, the research hotspot, technical challenge and future development direction of the PL method are discussed.
下载:
