【作者】 张浩；

【导师】 张庆君； 张过； Paco López-Dekker；

【作者基本信息】 武汉大学 ， 摄影测量与遥感， 2020， 博士

【摘要】 D-In SAR技术和时序In SAR技术是SAR领域最成功的应用技术之一,被广泛应用于从火山地形形变监测、滑坡变形监测、建筑物变形监测等领域。相比DIn SAR技术,时序In SAR技术有有效减小失相关因素影响、监测精度高等优点,本文在前人研究的基础上,提出了一种新的时序In SAR技术-超分辨率永久散射体干涉测量技术(Super Resolution Persistent Scatterer Interferometry,SR-PSI)。本文围绕SR-PSI技术展开研究,主要研究的目标是通过合适的策略将超分辨率处理技术融合到PSI技术中,建立一套完整的SR-PSI的处理流程,并通过仿真和真实数据验证SR-PSI的有效性。围绕此研究目标,本文做的主要工作如下:1.本文分析了当前时序In SAR技术中存在的问题。为了提高PSI中PS点的密度,本文提出了SR-PSI技术,将基于谱估计的SAR影像超分辨率重处理技术融入到PSI技术中。2.本文针对超分辨率重处理技术的原理、实际应用中的参数设置、快速算法等问题展开了研究,并研究了TOPS成像模式的超分辨率重处理问题。3.针对基于对角线加载的稳健Capon算法会加重重处理影像的边缘效应的问题,提出了修正的稳健Capon算法,基于Terra SAR-X数据实验结果表明,该算法能够有效减弱SAR影像重处理的边缘效应。4.针对超分辨率重处理影像分辨率自适应的特点导致传统PSC提取方法难以直接应用的问题,本文提出了基于顶点探测的PSC提取技术,并进一步研究分析了SR-PSI的整个技术流程。为了对SR-PSI的性能进行分析,我们研究了其评价指标,并进一步基于仿真数据和Terra SAR-X、哨兵1号TOPS数据进行了实验验证。仿真数据和真实数据实验结果均表明,SR-PSI优于传统的PSI处理流程,在SR-PSI中单时相Capon算法表现最优。对真实数据提取的PS点的形变速率、高程、形变标准差的对比分析表明,SR-PSI能够保持与传统PSI相同的相位质量。相比传统PSI技术,SR-PSI能够将PS点的密度提高30%-50%。5.针对超分辨率算法效率低的问题,本文提出了基于同质区域识别的SR-PSI和传统PSI联合处理的方法。首先推导了不同显著水平下的识别同质区域的变异系数阈值,然后分析了联合处理的技术流程。最后通过Terra SAR-X真实数据的实验结果表明,当变异系数阈值设置为0.75时,联合处理算法相比传统PSI算法能够将永久散射体的密度提高39%,而相比SR-PSI计算效率提高了48%,说明联合处理方法能够保持SR-PSI的有效性并有效提高其处理效率。本文通过以上研究详细分析了所提出的SR-PSI技术,并通过仿真数据和真实数据实验证实,相比传统PSI技术,本文提出的SR-PSI能够有有效提升PS点的数量,减少旁瓣PS点。希望本文的研究能够为相关研究提供一定的启发和帮助,并进一步应用于形变测量的工程实践中。更多还原

【Abstract】 D-In SAR and time-series In SAR are one of the most successful technologies in the filed of SAR,which have been widely used in many applications including volcanic terrain deformation monitoring,landslide deformation monitoring,building deformation monitoring.Compared with D-In SAR,time-series In SAR reduces the decorrelation and improve the accuracy.Based on previous studies,this paper proposes a new time-series In SAR chain-Super Resolution Persistent Scatterer Interferometry,SR-PSI.In this study,we focus on SR-PSI.We set the main objective of this research as to construct a full chain of SR-PSI by integrating the super-resolution reprocessing algorithms into PSI,and to validate the effectiveness of SR-PSI through both simulation and real-life SAR data.Our contributions are presented as follows.1.We analyze the problems of current time-series In SAR algorithms.In order to improve the density of PS points in PSI,we propose the workflow of SR-PSI,which integrates spectral-estimation-based SAR reprocessing algorithms into PSI.2.We study the theory of super-resolution reprocessing algorithms.Besides,we analyze several aspects in the implementation,and the fast algorithms.We also study the super-resolution reprocessing of TOPS data.3.In order to solve the problem the diagonal-loading based robust Capon algorithm may aggravate the edge effect,we propose a revised and robust Capon algorithm.The results of Terra SAR-X show that the algorithm can effectively reduce the edge effect of the reprocessed SAR image.4.The feature of adaptive resolution for the reprocessed SAR images causes the tradition PSC selection method can not be applied directly.We propose the peak-detection based PSC selection method,and present the full chain of SR-PSI.In order to analyze the performance of SR-PSI,we studied its evaluation workflow,and further carried out experimental verification based on simulation data and real-life data(Terra SAR-X,Sentinel 1 TOPS data).Both results of simulation data and real-life data show that SRPSI prevails the traditional PSI.The single-epoch Capon algorithm provides the best results in the context of PSI.We compare the the deformation rate,height,and standard deviation of PS extracted from real data and confirm that SR-PSI can preserve the phase quality as the traditional PSI.SR-PSI can increase the density of PS by 30%-50% than the traditional PSI.5.In order to improve the efficiency of SR-PSI,we propose to combine the SR-PSI and the traditional PSI by identification of homogeneous areas.We first derive the threshold of the coefficient of variation at different significance levels,and then analyzed the workflow of the combination processing.The experimental results of Terra SAR-X real data show that when the coefficient of variation threshold is set to 0.75,the combined processing algorithm can increase the PS density by 39% compared with the traditional PSI algorithm,and can improve efficiency of SR-PSI by 48%,which indicates that the joint processing method can maintain the effectiveness of SR-PSI and improve its efficiency.This paper analyzes the proposed SR-PSI technology in detail through the above research,and confirms its effectiveness through simulation data and real data experiments.Compared with traditional PSI,SR-PSI can increase the number of PS and reduce side lobe PS.We hope that the research in this paper can provide inspiration and help for related research,and be further applied to the engineering practice in deformation monitoring.更多还原