Modeling Of Time Series Radar Interferometry By Multi-Constraints And Extracting Of Ground Deformation

Ground deformation is a typical geological disaster caused by natural and human factors, which will hazard the safety of buildings, underground pipeline and transportation facilities. Therefore, it is necessary to monitor the ground deformation with high precision and effective technology. In the deformation monitoring, it is important to achieve the temporal and spatial changes rules, and the distribution characteristics of the deformation.Due to short revisit period, wide coverage and high precision, differential synthetic aperture radar interferometry (DInSAR) and persistent scatterers interferometry (PSI) have been widely used in ground deformation monitoring in recent years. However, existing methods achieve the deformation information for target areas generally only by single SAR data. By these methods, only one-dimensional deformation information in the direction of the radar line of sight (LOS) can be obtained and the horizontal displacement information is ignored. This will cause inaccurate vertical deformation extraction, and does not consider the influence of the horizontal displacement on the building structure, and pipeline facilities. In order to avoid the errors caused by the single observation and to better describe the surface deformation, multi-dimensional surface deformation monitoring is needed.In view of this, many multi-dimensional time series deformation calculation methods have been developed by domestic and foreign scholars, and widely applied in the field of volcano, landslide and so on. The representative MSBAS can simultaneously solution the vertical and east-west deformation time series and has a high applicability for different types of ground deformation. However, the observation equations have the problem of rank deficient and ill-posed, resulting in time series unstable and the solution non-unique.Hence this thesis proposes a multi-constrained time series InSAR (MCTS-InSAR) with the dual constraint of deformation model and regularization to achieve the correct deformation recovery. The main contents of this thesis are described as follows:(1) According to the problems of the SAR imaging radar interferometry technique in the surface two-dimensional time series monitoring, the thesis introduced a joint multi-source SAR satellite data for modeling. The vertical and east-west time series and deformation rate can be simultaneously extracted, and the problem of rank deficient and ill-posed can be effectively solved.(2) In order to ensure the reliability of the unwrapping phase in the time series model, the thesis developed a spatio-temporal three-dimensional phase unwrapping method. The phase unwrapping results show that three-dimensional phase unwrapping is smoother than the traditional two-dimensional spatial minimum cost flow (MCF) and the phase gradient jump variables of the three-dimensional phase unwrapping is significantly lower than MCF. In addition, compared with MCF, this method can increase the accuracy by 4 times.(3) In order to solve the problem of rank deficient and ill-posed involved in the two-dimensional time series observation solution, the thesis presents a two-dimensional deformation time series method (MCTS-InSAR) with dual constraints developed by the deformation model and ridge estimation regularization. A combination of linear and nonlinear deformation (such as periodic) is used to conduct the deformation transformation between the different satellite platforms, and the regularization is adopted to increase the observation equations so as to solve the rank deficient and the ill-posed of the observation equations. The real temporal and spatial baselines of the multi-platform are used to simulate the unwrapped differential interferometric phase. The reliability, applicability and accuracy of the MCTS-InSAR are determined by three factors, i.e. the time overlap, the deformation type and the noise. Simulation results show that the accuracy of MCTS-InSAR can reach mm level, and the accuracy of vertical direction is higher than east-west direction.(4) This thesis introduces a variance component estimation method to optimize the weights of the differential interference of the different satellite platforms so as to improve the reliability and precision of the vertical and east-west time series with MCTS-InSAR. The ascending orbits COSMO-SkyMed and the descending orbits TerraSAR-X imagery covering Shanghai urban area from 2009～2010 are used as test data. Results show that the vertical deformation rate is up to-19 mm/year and east-west is up to-13 mm/year in Hongkou district. The results show that the accuracy of the vertical and east-west mean rate can reach mm level. The priori and variance estimation of the unit weight error show that the maximum and the minimum improvements are 5.84 mm and 0.90 mm respectively, and the accuracy of time series deformation can be improved by 2.17 mm.In summary, this thesis aims to better reconstruct the surface vertical and east-west time series and accurately achieve mean deformation rate. To ahieve this, a MCTS-InSAR method is developed in this thesis and experimental evaluations for this method has been carried out. Experimental results show that this method can reach the accuracy of mm level and effectively solve the rank deficient and ill-posed problem of the observation equation.