【作者】 陈思；

【导师】 陈宇；

【作者基本信息】 中国矿业大学 ， 摄影测量与遥感， 2023， 硕士

【摘要】 城市地区常因基础设施建设、地下水和矿产资源开采等产生不同程度的地表形变,威胁着居民的生命财产安全。因此,定期进行城市地表形变监测对预防相关地质灾害、促进城市可持续发展具有重要意义。传统测绘手段,如三角高程测量、水准测量、全球卫星导航系统(Global NavigationSatellite System,GNSS)技术等,因其高精度的特点被广泛应用于地表形变监测领域,但其离散点监测的方式有时不能满足广域地表形变监测的需求。近年来发展的时序InSAR(Interferometric Synthetic Aperture Radar,InSAR)技术常用于获取长时序地表形变监测,在保证空间分辨率的基础上可获取长时间序列地表形变信息,有助于刻画和揭示地表形变时空演变规律。然而,城市地区地表形变量级较小和连续缓慢的特点,使得该特征区域对大气延迟等误差和噪声的精细估计与控制尤为重要。现有城市地表形变时序InSAR监测应用研究中多采用时空滤波方法去除大气延迟误差和噪声信号,存在引入新噪声或形变信息被误判为噪声被去除等问题。针对上述问题,本文以时序InSAR技术为基础,提出了融合主成分分析法(Principal Component Analysis,PCA)和变分贝叶斯独立成分分析法(variational bayesian Independent Component Analysis,vb ICA),旨在提取高精度地表形变时空信息。论文研究内容和成果如下:(1)阐述了InSAR/D-InSAR技术的理论原理和研究发展现状,分别介绍了永久散射体干涉测量(Persistent Scatterers InSAR,PS-InSAR)技术和小基线集干涉测量(Small Baseline Subset InSAR,SBAS-InSAR)技术的基本原理;同时分析了大气对雷达信号的延迟效应。(2)本文以66景覆盖徐州地区的Sentinel-1A影像为数据源,采用时序InSAR技术对典型地质灾害易发城市徐州进行地表形变时空监测。提出了融合PCA/vb ICA和时序InSAR的时空特征挖掘与分析方法,通过对时序InSAR信号主成分的时空分析,构建时序地表形变模型,结合多项式函数,实现误差及噪声信号的分层估计,进而提取高精度、小量级的地表形变信息。使用CORS站点数据和水准数据验证两种方法得出的地表形变信息的精度,表明两种方法得出地表形变的准确性;并对比原始差分干涉图、时空滤波后干涉图、SAR影像过境时的GACOS数据改正后的干涉图及本方法改正后的干涉图及他们形变结果的RMS,获得两种方法相较于其他方法的精度提升程度,最后将PCA法得出的时空形变结果和vb ICA法得出的时空形变结果对比。(3)结果表明徐州市地表形变主要分布在城区、地铁沿线及采煤塌陷区,2018～2022年的形变速率约为-17～35 mm/a;近8年内,徐州市区内出现多处因城市建设引发的沉降区,采煤塌陷区地表表现出不同程度的次生形变;PCA方法与vb ICA方法得出的地表形变结果大致相同,局部地区有差异,且vb ICA法比PCA法具有更高精度。更多还原

【Abstract】 Urban areas often experience varying degrees of surface deformation due to infrastructure construction,groundwater use,and mineral extraction,and these deformations can threaten the safety of residents and their properties.Therefore,regular urban surface deformation monitoring is of great importance for preventing geological hazards and promoting sustainable urban development.Traditional surveying methods,such as triangulation,leveling,and Global NavigationSatellite System(GNSS)technology,are widely used in the field of surface deformation monitoring due to their high accuracy,but their discrete point monitoring mode sometimes cannot meet the requirements of large-area surface deformation monitoring.In recent years,the timeseries Interferometric Synthetic Aperture Radar(InSAR)technique has been commonly used to obtain long-term surface deformation monitoring data.This technique can acquire long-term surface deformation information while ensuring spatial resolution,which is helpful in depicting and revealing spatio-temporal evolution laws of surface deformation.However,the small scale and continuous slow deformation characteristics of surface deformation in urban areas make it particularly important to accurately estimate and control errors and noise such as atmospheric delay.In the existing urban surface deformation InSAR monitoring application research,spatio-temporal filtering methods are often used to remove atmospheric delay errors and noise signals,but there are still problems such as the introduction of new noise or the removal of deformation information that is mistakenly judged as noise.To address these issues,this article proposes a method that combines Principal Component Analysis(PCA)and variational Bayesian Independent Component Analysis(vb ICA)based on time-series InSAR technology.The aim is to extract high-precision spatio-temporal information on surface deformation.The research content and achievements of this thsis are as follows:(1)This thsis elaborates on the theoretical principles and research development status of InSAR/D-InSAR technology for deformation measurement.It introduces the basic principles of two techniques,namely Persistent Scatterers InSAR(PS-InSAR)and Small Baseline Subset InSAR(SBAS-InSAR),and analyzes the delay effect of the atmosphere on radar signals.(2)This article uses Sentinel-1A images covering 66 scenes as the data source to carry out spatio-temporal monitoring of surface deformation in Xuzhou,a city prone to typical geological disasters.The thsis proposes spatio-temporal feature mining and analysis methods that combine PCA,vb ICA,and time-series InSAR.By analyzing the spatio-temporal characteristics of the main components of the time-series InSAR signal and combining it with a polynomial function to realize layered estimation of errors and noise signals,high-precision and small-scale surface deformation information is extracted.The accuracy of surface deformation information obtained from the two methods is verified using CORS station data and leveling data.By comparing the RMS of the deformation results obtained by two methods with those obtained by other methods,the accuracy improvement of the two methods compared to other methods is obtained.Finally,the results obtained by the PCA method and the vb ICA method are compared.(3)The results show that Xuzhou is mainly distributed in urban areas,subway lines,and coal mining subsidence areas.The deformation rates from 2018 to 2022 were approximately between-17～35 mm/a.In recent years,there have been multiple subsidence areas in the city caused by urban construction,and the coal mining subsidence areas have shown varying degrees of secondary deformation;The surface deformation results obtained by PCA method and vb ICA method are roughly the same,with differences in local areas,and vb ICA method has higher accuracy than PCA method.更多还原