Deformation monitoring of single buildings using meter-resolution SAR data in PSI

Kurzfassung

In this paper the feasibility to monitor the shape and deformation of single buildings from space is investigated. The methodology is based on a fusion of persistent scatterer (PS) point clouds obtained from several stacks of meter-resolution synthetic aperture radar (SAR) data. This kind of high resolution imagery as well as accurate orbit information is available from, e.g., TerraSAR-X. The stacks are processed individually applying persistent scatterer interferometry (PSI), which provides deformation and height estimates for the PS. However, the geocoded PS point clouds cannot be simply merged in one common coordinate system, like UTM, by reason of residual offsets with respect to their final true positions. These deviations originate from the height uncertainty of the reference point, which has to be chosen during PSI processing of each stack. The presented methodology allows for a fusion of several PS point clouds, i.e., the correct reference point heights can be recovered. The algorithm is based on a point cloud matching procedure that consists of a determination of appropriate point correspondences and a minimization of the distances between all selected pairs of points in a least-squares sense. In addition, the reconstruction of the original motion vector from the deformation measurements in line of sight provided by PSI is desirable. The availability of separated motion components in vertical and horizontal directions greatly enhances the insights into deformation events at buildings and ground. To this end, the fused point clouds are used for a decomposition of motion components. By reason of the limited sensitivity of ascending and descending stacks of TerraSAR-X to deformation in north-southern directions the reconstruction of motion vector components is restricted to components in west-eastern and vertical directions. The latter are estimated in a least-squares adjustment including all PS within a spatially limited area. Deformation estimates of stacks from ascending and descending tracks must be included in order to separate motion components. The latter cannot be determined precisely from a combination of solely equal heading tracks, as the line of sight does not differ enough. Finally, deformation maps of the urban area are available that separately show seasonal and linear deformation in horizontal as well as vertical directions. These maps comprise important information on subsidence or uplift as well as structural stress at buildings due to thermal dilation. As a result of the presented methodology (and for the first time) sufficient and precise motion estimates are available for a detailed monitoring of single objects using meter-resolution SAR data in PSI. Several examples are discussed using the results of motion component estimation based on the fusion of four data stacks evaluated by PSI.