| As a newly arisen space geodetic technique, synthetic aperture radar interferometry (InSAR), is well known as one of the most effective and powerful tool for monitoring ground deformation, which has series of advantages (e.g. higher precision, larger monitoring field, unlimited to weather condition). Nowadays, InSAR is widely used in research area of volcano, earthquake, glacier, landslide and subsidence. In recent years, to mitigate the negative effects of two major limitations in conventional InSAR for deformation tracking, i.e., spatiotemporal decorrelation and atmospheric artifacts, relevant researches have focused on the multi-temporal InSAR based on time series of SAR images for analyzing the temporal deformation behavior. Along with other algorithms have been also proposed to adapt to the specific applications in last ten years, an available systematic theory named persistent scatterer interferometry (PSI) is completed finally. However, practical application indicated that some problems exist in the key modeling and processing procedures of PSI, such as observation network construction, deformation series modeling and3D modeling estimation, e.g. Thereby, this thesis focuses on investigating basic models and algorithms of networking PSI to search an available solution. As the fundamental theory of PSI time series analysis, this paper analyzes the components of interferometric phases, the principles of DEM generation and deformation exploration. In addition, polynomial modeling of systematic errors and relevant solution for controlling error propagation are discussed also. For validation purpose, a series of experiments have been implemented. The research results indicated that the polynomial modeling method and repeated iterative operation algorithm is effective for improving the precision and reliability of InSAR surveying. Along with the application of the new generation spaceborne SAR systems (e.g., TerraSAR-X and COSMO-SkyMed data) in recent years, the spatiotemporal resolution of image data has been improved apparently. With the image data of1meter in pixel spacing and the cycle of a few days, there will be more PS targets can be detected and applied in PSI time series analysis, which makes a good condition to monitor ground deformation more amply and accurately. However, facing to so mass observation data, the lengthy time spending and possible data overflowing cannot avoid. Therefore, for the purpose of the efficient PSI time series analyzing for large scale and long term with enough accuracy and reliability, it is necessary to improve the conventional networking differential phase observation model and relative algorithm. To balance between computational efficiency and solution accuracy, this paper presents a hierarchical approach of PS networking and solution for deformation analysis using time series of high resolution satellite SAR images. Its basic strategy is constructing global control network and a series of local triangulated irregular network (TIN) hierarchically. Because the image space has been divided into several subregions, there are relatively few PSs exist in each subnetwork. Thus the time consuming of the LS solution can be reduced in the way of geometric series. While global control network, constructed by simplified model though, ensures the internal geometric relationships of the image space. For validation purpose, the northwestern part of Tianjin is selected as the testing area, and40TerraSAR-X images collected over this area between2007and2010are applied in PSI time series analyzing, based on TIN and hierarchical network respectively. The comparison with the ground-based leveling results indicates that the linear deformation rates, derived by different network models, have the common accuracy, which can reach up to a millimeter level. While the time consuming and memory used of the LS solution with the hierarchical model is markedly less than TIN, the ratio are only22%and0.8%, respectively. This indicates that the hierarchical model and relevant LS solution raises the calculation efficiency of the PSI time series analyzing (for high resolution SAR imagery particularly) availably.In terms of PSI solution based on time series of SAR images, current PSI modeling is basically in accordance with statistical analyzing to linear velocity and nonlinear deformation components, respectively. This model is enough to fit the slowly accumulated ground displacement. But for uneven deformation or seasonal cycle trend, other statistic models (e.g., hyperbola function,2th/3rd order polynomial and sine/cosine trigonometric function), are more suitable. Because of complex interaction effects due to multiple natural and human factors, the real ground deformations always reflect mixed diversification characteristic. Therefore, it is not enough to simulate real displacement with a single statistic model. However, the time series observation based unknown component solution cannot avoid the solution space’s divergence. Thus it is impossible to exploration accurate information though higher order polynomial modeling. To solve the contradiction between limitation of single model and solution problem of multiple unknown parameters, this paper proposes an integrated statistic model, which mixed high order polynomial and trigonometric function together. Moreover, an iteration process algorithm for solving multiple unknown components is provided in this thesis also. For validation purpose, the northwestern part of Tianjin is selected as the research area, and40TerraSAR-X images collected over this area between2007and2010are applied in PSI time series analyzing, based on proposed integrated statistic modeling and iteration process solution. The comparison with the ground-based leveling results indicates that the ground displacements, derived by proposed model and algorithm, reach up to a millimeter accuracy level. In addition, the comparative analysis among multiple deformation components indicated that this integrated statistic model is suitable for fitting the complicated ground displacement induced by interaction effects of multiple natural and human factors.PSI analysis with time series of SAR images collected by a single side-looking mode from a single satellite platform can only reveal one-dimension (1D) ground deformation along radar line of sight (LOS). This paper proposes the modeling and algorithm for extracting three-dimensional (3D) deformation velocity field using a multi-platform PSI method. Its basic strategy includes two parts. The SAR images collected by each single platform are first used to estimate LOS deformation velocity field by PSI analysis, and then all the1D deformation velocity fields obtained from different platforms are combined to resolve3D deformation velocity field by least squares solution. For validation purpose, a small area located the northwestern part of Tianjin is selected as the testing area, and39TerraSAR-X images,23ENVISAT ASAR and16ALOS PALSAR images collected over this area between2007and2010are used to estimate the vertical deformation velocity field and the horizontal deformation velocities. Comparison with the ground-based leveling results and the existing GPS analysis indicates that the vertical deformation velocities derived by the combined solution can reach up to a millimeter accuracy level, and the horizontal deformation velocities are in good agreement with those derived by GPS analysis. The proposed multi-platform PSI analysis can be used to calibrate the1D deformation velocity field and to reconstruct the3D deformation velocity field. External reference data is needless in this model and solution, which is another contribution of this3D multi-platform model. |
PDF Full Text Request