High Precision 3D Localization and Motion Analysis of Persistent Scatterers using Meter-Resolution Radar Satellite Data

Kurzfassung

In this thesis the impact and benefits of high resolution synthetic aperture radar (SAR) data for persistent scatterer interferometry (PSI) is investigated. The data is obtained by the German remote sensing satellite TerraSAR-X. From the high resolution compared to available SAR satellites so far - like ERS or ENVISAT - a large increase in numbers of persistent scatterers (PS) is expected. Therefore, the influence of different acquisition parameters on the appearance of PS is investigated in order to analyze the opportunistic character of these points. It is shown that the geometrical configuration of the sensor with respect to the objects on ground has to be considered prior to the ordering of data, as the local PS densities vary between 40,000 and 100,000 PS/sqkm in the analyzed areas. The different parameters are discussed in order to help with the choice of assumed best configuration, depending on the goal and demands of deformation monitoring for the area of interest. Several deformation phenomena on single buildings within two test sites of Berlin and Munich, Germany, are presented and discussed. The benefits for single object observation due to the high density of PS is demonstrated and the high precision of linear as well as periodic displacement is revealed. The availability of several stacks of the same area allows for an interpretation of motion directions with respect to horizontal as well as vertical directions. Very interesting seasonal patterns are discovered at Berlin central train station and the origin of the motion is shown to be related to thermal expansion of the steel construction. Ground truth inspections reveal special constructional devices that enable non-destructive motion at the magnitudes of the observed deformation. Furthermore, the localization precision of point scatterers is investigated by a comparison of theoretical accuracies with real data results of PSI. The regular patterns of PS on facades of modern buildings are used to derive the standard deviations of PS positions, which fit the theoretical assumptions of 2 cm in range, 4 cm in azimuth and 60-140 cm in elevation. By reason of the limited variation of orbital positions the location in elevation direction is available only at a precision by a factor of 20-30 worse than for azimuth or range direction. This is confirmed by the evaluation of PS at several facades of buildings. Based on the precise 3D localization of PS, a new methodology for a fusion of independently geocoded PS point clouds obtained from stacks of high resolution SAR data is presented. The goal of this approach is a representation of the mapped objects by preferably completed 3D point clouds. It is suitable for a combination of stacks of arbitrary track directions, i.e., especially the gain in information of fusing ascending and descending stacks is accomplished. The unknown relative height offsets of the reference points are solved without the usage of any reference data within the area. The algorithm is based on a least-squares adjustment of selected pairs of PS assumed (but not necessarily) to be identical. After the fusion the majority of buildings are covered by PS on all sides. Hence, single object observation is feasible. The combination of stacks of opposite viewing directions additionally allows for a calculation of motion components in vertical and horizontal directions. A straightforward algorithm is presented and results are provided for one of the test sites. The findings of the initial interpretation of motion directions by a visual comparison of PSI results of different stacks are confirmed and further examples of seasonal displacement in urban areas are provided.