Vertical Structure Information Retrieval Based On Multi-baseline SAR Data At P-band

As an active microwave remote sensing technology,Synthetic Aperture Radar(SAR)has been widely used in resource surveys,topographic mapping,disaster monitoring and environmental assessment,etc.,due to its all-day,all-weather working capability and penetration ability to certain features,which has now become one of the most advanced earth observation technologies in the world.In particular,benefitting from the strong penetration capability to surface and subsurface features,SAR systems working at P-band can provide the internal backscattering information that cannot be obtained by high frequency SAR systems,hence are now increasingly favored by relevant experts,and have shown great potential applications in forest vertical structure retrieval,subsurface targets detection,etc.Among these systems,the most concerned one is the BIOMASS mission proposed by European Space Agency(ESA)in 2013,which will be launched in 2022.This mission will carry,for the first time in space,a fully polarimetric P-band SAR system,both of polarimetric SAR interferometry(Pol In SAR)and SAR tomography imaging technologies will be used for mapping global forest biomass,for the purpose of providing important scientific evidences for studying the internal mechanisms and connections between carbon cycle and global climate change.On the basis of ESA-MOST Dragon-4 Programme,this thesis is committed to the development of vertical structure parameter inversion algorithms for forest and glacier scenarios using P-band multi-baseline SAR data,by taking advantage of the multiple datasets obtained from several airborne flight tests that were carried out in the early demonstration stage of the BIOMASS satellite.The main contents and findings including the following aspects:(1)The strong penetration capability of P-band SAR system makes the signals in all polarizations to be contributed by significant ground scattering,thus the classic three-stage inversion algorithms can become problematic as the volume-only coherence cannot be obtained.To solve this problem,we propose a new parametric height inversion approach,where forest canopy height can be obtained by minimizing the least-square problem between random volume over ground(RVo G)model predictions and multibaseline SAR data at P-band.On this basis,we furtherly compare the inversion accuracy between this method and another one,namely the forest height estimation approach by directly thresholding tomographic profiles,to demonstrate the effectiveness of the proposed method.Experiments results have shown that both algorithms were observed to provide reliable results for mature forest stand,with an accuracy superior than 3 m,which can meet the accuracy of measurement requirements.(2)On the basis of multiple flight tracks from slightly different looking angles,SAR tomography technique allows a conversion of multibaseline SAR data to multilayer images,where each image contains the forest backscattered information of a certain height in vertical direction.Above ground biomass(AGB)inversion can then be obtained by constructing the linear regression models between tomographic images and in-situ biomass.However,the propagation of transmitted wave through the volume layer can be attenuated by the physical effect of wave extinction,and the presence of this factor can weaken the relationship between tomographic backscatter power and biomass.On this basis,in this thesis we propose to correct tomographic data by accounting for the effect of extinction occurred in wave propagation.Experimental results have shown that the linear coefficient and slope between tomographic backscattered images and in-situ AGB improve dramatically by considering the role of wave extinction.Moreover,the accuracy of retrieved biomass is also improved.(3)The long-wave characteristic of P-band SAR data enables to retrieve the internal scattering information of subsurface.However,refraction phenomenon occurs when radar wave penetrate the interface between atmosphere and subsurface,leading to the change of propogation path and wave velocity.If this is not correctly accounted for in the focusing process,the resulting tomographic images are likely to appear distorted,hindering the interpretation of the results.In view of the above problem,this thesis has proposed a new method for positioning and correcting the 3D structure of subsurface objects.Based on carrying out simulation experiments and real data experiments,the precise location of the targets beneath ice surface and the quantitative estimation of the true penetration depth are realized,thus verifying the effectiveness of the proposed method in this paper.