Interferometric Phase Optimization In Tspol InSAR Surface Deformation Monitoring With Consideration Of Statistical Properties

Interferometric Synthetic Aperture Radar(InSAR)technology is a newly important means of image geodesy,which has the unique advantages of rapid,large-scale,and all-weather surface deformation monitoring.As a direct observation of InSAR technology,the interferometric phase and its quality determine the feasibility and affect the accuracy of surface deformation monitoring.Therefore,how to extract the reliable systematic phase series from InSAR temporal data,namely,interferometric phase optimization research,is an important premise to ensure the universality of InSAR technology and high-precision surface deformation monitoring.Currently,the interference phase optimization algorithms are facing the higher challenge of high-dimensional random signal processing and has been significantly limited by the fast decorrelation in natural scenarios.With the flourishing development of spaceborne Polarimetric SAR platform,to improve the density and deformation monitoring accuracy of the measurement points,many scholars and institutions are inspired by the polarimetric InSAR(PolInSAR)theory and introduce multi-polarimetric scattering information of a ground object into the time-series InSAR(TSInSAR)data configuration,called the time-series PolInSAR(TSPolInSAR)technology.However,the current TSPolInSAR deformation monitoring technology still cannot effectively solve the above technical limitations.There are two main reasons: 1)to utilize of polarimetric scattering information only partially,which cannot accurately evaluate the relevant properties of scatterers;2)to focus on the artificial targets with stable scattering behaviors,which cannot be applied to the highly dynamic changing natural scenarios.In view of this,this thesis takes Persistent Scatterer(PS)and Distributed Scatterer(DS)as the research objects,fully utilizes multidimensional information from polarization,time,and space domains,starts from statistical sample enrichment,random consistency measure,and complex least square adjustment based on the idea of the information enhancement,and systematically studies the accurate identification of scatterer,polarimetric coherence optimization,and systematic phase series reconstruction.Then,a complete new TSPolInSAR surface deformation monitoring method with emphasis of the interferometric phase optimization has been established to realize the accurate estimation and quality control of high-dimensional random signals,opening a new way for the high-precision surface deformation monitoring.The main contributions and innovations of this thesis are as follows:(1)A new PS-oriented interferometric phase optimization considering time-series polarimetric stationarity and scattering amplitude optimization is proposed,which solves the problem of wrong selection of non-highly coherent points in traditional polarimetric PS interferometry(PSI)technology and enhances the reliability of subsequent observation data.The traditional polarimetric PSI techniques take the amplitude dispersion index as the optimization criterion and evaluate the time-series amplitude stationary of each polarimetric channel to find an optimal polarization channel.However,due to the unstable statistical characteristics of quality indicators,many lowly coherent points are mistaken as PS candidate points.To reduce the adverse effect on the subsequent deformation results,this thesis integrates the following two innovations into the existing PSI technology and proposes a novel polarimetric PSI deformation monitoring method: in terms of the identification of highly coherent scatterers,the temporally polarimetric stationarity of scatterers is evaluated by the trace moment statistical properties of the time-series polarimetric coherence matrices;in terms of interferometric phase optimization,the PolInSAR amplitude optimization method is extended to the single-reference interferometric framework,which is beneficial to suppress the signal-to-noise ratio decorrelation.Real data experiments in the Los Angeles city of the United States show that the new method not only improves the accuracy of deformation monitoring,but also increases the density of measurement points to nearly twice that of the traditional method and shortens the calculation time to less than1/200 of the traditional method.(2)A DS-oriented non-local(NLM)homogeneous pixels filtering method based on homogeneity measurement is developed,which considers the spatial similarity and set homogeneity of multidimensional SAR data and innovatively realizes the complex least squares adjustment-based heterogeneous pixel removal method.At present,the widely used NLM filtering method still faces two key problems: first,it is challenging to construct a shape-adaptive patch,which leads to the homogeneous pixels omission problem;second,the mainstream Lee estimator achieves the balance estimate between the central pixel value and the mean value,which cannot eliminate the wrongly selected heterogeneous pixels.To overcome the limitations mentioned above,this thesis proposes a homogeneity measure-based non-local filtering framework for(Pol)(In)SAR images and fully proves the effectiveness of the following innovations with simulation data experiments: firstly,to ensure the sufficiency of homogeneous pixels selection,the shape-adaptive patch is selected from multiple preset patches with different shapes based on the likelihood ratio test and multi-patches matchings method;secondly,based on complex least squares adjustment,the iterative re-weight estimation and the homogeneity measure are combined to adaptively remove the wrongly selected heterogeneous pixels from the selected pixel set.Real experiments in Flevoland farmland and Spanish area demonstrate the advantages of the proposed method over traditional methods in many aspects such as speckle noise suppression,texture structure enhancement,point target preservation,and coherence amplitude estimation.(3)Introducing the innovative concept of polarimetric statistical samples,a DS-oriented polarimetric coherence optimization method based on the time-series total power coherence matrix construction is proposed,which breaks the mindset of the traditional polarimetric optimization method and improves the calculation efficiency by 6orders of magnitude while greatly improving the phase quality.The DS-oriented polarimetric coherence optimization methods mainly take the temporally averaged interferometric coherence amplitude as the optimization objective and search for an optimal polarization channel in a given polarization space to improve the quality of the time-series interferometric phase.However,due to the existence of the finite sample effect,the complex coherences obtained by the traditional methods above tend to be biased from the noise-free one,and hence the optimization performance is unstable,and the computation cost is huge.Different from the traditional methods,this thesis proposes a novel approach to utilize the multi-polarimetric scattering information of a ground object: consider the multi-polarimetric channels as statistical samples and perform the polarization stacking operator to construct the time-series total power coherence matrix.In the case of PolInSAR data,it has been demonstrated that the newly constructed total power complex coherence has rich theoretical values for different types of ground targets.In the case of TSPolInSAR data,simulation experiments show that the time-series interferometric phase optimization accuracy of the proposed method is48.6% higher than that of the traditional method;real data experiments in San Francisco,USA,show that compared with the traditional method,the new method improves the interferometric phase quality and clears the phase fringe,and improves the calculation efficiency by 6 orders of magnitude.(4)Proving the feasibility of small baseline phase linking theoretically,a DS-oriented time-series interferometric phase adaptive reconstruction method based on interferometric pair selection optimization and adaptive regularization is proposed,which eliminates the over-dependence of traditional small baseline method on the temporal baseline threshold and solves the worsen ill-posed problem in the parameter solution.Based on the fast exponential temporal decorrelation model,it has been proved theoretically that although the only utilization of highly coherent interferometric pairs can effectively improve the time-series phase optimization accuracy,at the same time the positive definiteness of the time-series coherence matrix is further destroyed,making the maximum likelihood estimate more instable.In view of this,this thesis integrates the following two innovations into the maximum likelihood estimation and proposes an adaptive phase linking method: in terms of interferometric pairs selection optimization,faced to the problem that the traditional methods exceedingly depend on the temporal baseline threshold,a consistency measure method based on the prior and posterior variances of the coherence magnitudes is proposed to gradually remove the candidate subset of lowly coherent pair and construct a suitable small baseline network;in terms of parameter solution,faced with the deteriorated regularization matrix,this thesis considers that the fitting error can reflect the stability of the regularization solution and proposes an adaptive regularization solution algorithm based on the optimal selection of damping factors.The simulation experiments show that the new method improves the time-series phase optimization performance and the deformation rate estimation accuracy by 51.3% and 73.2% than the traditional method,respectively.Real experiments in Shanghai Pudong International Airport show that compared with other traditional methods,the proposed method can not only clearly reconstruct the interferometric fringe with the temporal baseline of one year,but also effectively detect the high dynamic deformation characteristics of airport runways or construction land.(5)For lowly coherent vegetation areas with frequent landslides,we integrate the above four methods into the existing Squee SAR technical framework to jointly process both PS and DS targets,develop a TSPolInSAR surface deformation monitoring method with consideration of the statistical properties and apply it to the Xinpu landslide area in the Three Gorges Reservoir area.,which can improve the monitoring accuracy of surface deformation in vegetation area and enhance the reliability of interpretation of landslide deformation mechanism.Landslide disaster is one of the main types of natural geological disasters in China,often occurring in vegetation-covered areas.However,the reliability of the interferometric signal is significantly reduced in vegetation-covered areas due to the severe attenuation and temporal decorrelation of microwave signals in the vegetation canopy,and even it is prone to deformation monitoring blind areas.Aiming at the problem of TSInSAR landslide deformation monitoring in the vegetationcovered area,to fully excavates the interferometric information of both PS and DS targets in the SAR view,this thesis considers the time-varying polarimetric properties of vegetation and integrates the above innovative methods into the existing coherent scatterer interferometry(CSI)technology,and proposes a new polarimetric CSI technology for monitoring surface deformation in vegetation-covered areas.The wavelength-short SAR data experiments in the Xinpu landslide area of the Three Gorges Reservoir area with dense vegetation shows that compared to traditional methods,the deformation rates estimated by the new method are closer to the monitoring results based on wavelength-long SAR data;taking a measure point in the central deformation area of landslide as an example,the new method can accurately estimate the seasonal deformation sequence with a period of one year,whose deformation monitoring accuracy is improved by 30.2% compared with the traditional method.In addition,taking the GNSS data-based estimated results as the reference truth value,only the proposed method in this thesis successfully and quantitatively evaluates the deformation mechanism and time-delaying law of the Xinpu landslide driven by hydrological factors.