The Fusion of SAR Tomography and Stereo-SAR for 3D Absolute Scatterer Positioning

Kurzfassung

For decades, spaceborne Synthetic Aperture Radar (SAR) interferometry has evolved into a widely used geodetic method for three-dimensional mapping and analysing the geophysical processes of the surface of the earth. Nevertheless, effects such as atmospheric disturbances, temporal and geometric decorrelation are considered as the main degrading factors for the quality of Interferometric Synthetic Aperture Radar (InSAR) products. By overcoming the major deficiencies of the conventional InSAR, Persistent Scatterer Interferometry (PSI) proved the applicability of the interferometric methods also in monitoring of urban areas by restricting the analysis only on the time-coherent scatterers. In particular, the launch of modern SAR sensors, such as the German TerraSAR-X, characterized with very high spatial resolution and short revisit times has enhanced the capability of PSI in mapping and deformation monitoring of urban areas. However, PSI considers only a single dominant scatterer in each resolution cell which is not a valid assumption in urban areas due to the prevalent occurrences of layover phenomenon. This gives the motivation to exploit the most advanced SAR interferometric method, namely tomographic SAR inversion (TomoSAR) including SAR Tomography and differential SAR Tomography. TomoSAR coupled with very high resolution TerraSAR-X images produces the most detailed multi-dimensional maps of urban areas by distinguishing among multiple scatterers within a resolution cell. Nevertheless in TomoSAR, similar to conventional InSAR and PSI, elevation and deformation rates are estimated with respect to a previously chosen reference point which makes them relative 3D estimates. One unique feature of TerraSAR-X is the precise orbit determination and high geometrical localization accuracy of the sensor. After compensating for the most prominent geodynamics and atmospheric error sources, the absolute two-dimensional (range and azimuth) positions of targets such as corner reflectors and persistent scatterers can be estimated to centimetre-level accuracy, a method called SAR Imaging Geodesy. Moreover, using two or more SAR observations acquired from different satellite orbits, their absolute 3D positions can be retrieved by means of Stereo-SAR. In this thesis a framework is proposed that fuses the SAR imaging geodesy and TomoSAR approaches to obtain absolute 3D positions of a large amount of natural scatterers. The method is further utilized in order to automatically fuse multi-track TomoSAR point clouds. The methodology is applied on four Very High Resolution (VHR) TerraSAR-X spotlight image stacks acquired from the city of Berlin. The horizontal and vertical localization accuracy of the obtained fused point cloud is analysed by comparing them with a highly accurate geo-localized Digital Surface Model (DSM) of Berlin obtained from aerial laser scanning. The fusion of absolute TomoSAR point clouds obtained from cross-heading tracks also allows for the decomposition of the Line of Sight (LOS) motion observations into the real 3D motion vectors. As the final contribution of this study, the 3D motion decomposition of seasonal deformation observations is carried out on two test sites located in the city of Berlin.