| The new land formed by reclamation project on the estuary coast has the characteristics of high water content and large void ratio.The self-weight consolidation of the dredger fills will induce obvious land subsidence,which will cause damage to infrastructure such as roads and buildings,and even intensify the flooding caused by storm surges.Since Multi-temporal InSAR(MT-InSAR)has the ability to extract large-scale and high-precision surface deformation signal,it has been applied to the monitoring of surface deformation.However,the large-scale,slow,and unevenly distributed subsidence process caused by the self-weight consolidation of the dredger fills often lasts for several years or even decades.Limited to the service life and the on-orbit operation time of the satellite,the single Synthetic Aperture Radar(SAR)satellite platform cannot meet the requirements of monitoring and recording the complete consolidation subsidence process of the newly formed land at a long time period requirement.Effectively combination of the multi-satellite platform MT-InSAR time series can realize the requirement of regularity monitoring the newly formed land and analysis of the complete consolidation subsidence process.However,the uncertainty in the combination methods and optimization of combination strategy of the multi-satellite platform MT-InSAR deformation time series remains to be further studied.In addition,in order to further study the losses and impacts caused by surface deformation in the newly formed land,it is also necessary to effectively classify the land cover of the newly formed land.The multi-spectral remote sensing information,SAR backscattering intensity information,and interference coherence information can be used to improve the classification accuracy.However,in the newly formed land,the phenomenon of coherence abnormal reduction has occurred,which makes the same land cover has lower coherence in newly formed land compared to other areas.The phenomenon seriously affects the accuracy of multi-source remote sensing land cover classification while using coherence information.The purpose of this study is to analyze and determine the multi-platform MT-InSAR deformation time series combination optimization strategy,and analyze the uncertainty in the combination method.At the same time,this study analyzes the incoherence characteristics of the region with abnormally reduced coherence and eliminates the classification errors caused by the abnormal reduction of coherence to improve the accuracy of multi-source remote sensing land cover classification.The study area is the new formed area in the eastern part of Shanghai,which includes Pudong International Airport,Nanhui Dongtan and Lingang New City.In this thesis,based on 35 ENVISAT ASAR(ENV)images obtained from February 26,2007 to September 13,2010,61 COSMO-SkyMed(CSK)images acquired from December 7,2013 to March 18,2016 and 33 Sentinel-1A(S1A)images collected from July 08,2015 to August 26,2017,the deformation time series of three satellite platforms in three different periods are obtained by the Small Baseline Subset(SBAS)method.Based on the MT-InSAR deformation time series obtained by ENV,CSK and S1 A,the MT-InSAR deformation time series of a combination of three satellite platforms is realized by singular value decomposition(SVD)method,consolidation subsidence model,and nonlinear fitting method.For different multi-level combination strategies,this study discusses the differences between the combination deformation time series obtained by different multi-level combination strategies.The uncertainty of nonlinear fitting method based on consolidation settlement model is analyzed,and the nonlinear fitting optimization method of consolidation settlement model based on Bagging integrated learning is proposed.In terms of the influence of InSAR temporal incoherence effect on land cover classification,this study first discusses the phenomenon of abnormal coherence reduction in newly formed land and its impact on land cover.Then this study analyzes the influencing factors of abnormal coherence reduction in newly formed land.After that,this study uses the 33 S1 A images from 2015 to 2017 to construct the temporal incoherence model of the newly formed land,and finally introduces the temporal incoherence model of the newly formed land into the multi-source remote sensing land cover classification,and apply it to Pudong International Airport.The classification results show that the classification accuracy is 86% and the Kappa coefficient is 0.82.Compared with the classification results based on Sentinel-2 true color image and standard false color image,the overall accuracy increased by 22% and 11%,respectively,and the Kappa coefficient increased by 0.29 and 0.14,respectively. |
PDF Full Text Request