Advanced stacking techniques and applications in high resolution SAR interferometry

Kurzfassung

Interferometric Synthetic Aperture Radar (InSAR) is a satellite remote sensing technique that provides information about the topography and deformation of the Earth’s surface. In recent years, InSAR’s capabilities have been considerably improved with the launch of high resolution SAR satellites such as TerraSAR-X, TanDEM-X and COSMO-SkyMed. Mapping of urban areas and even single buildings is now facilitated via multitemporal InSAR techniques, for instance, Persistent Scatterer Interferometry (PSI) and SAR Tomography (TomoSAR). These methods exploit long-time coherent scatterers, e.g. the so-called Persistent Scatterers (PSs), and provide elevation and surface displacement measurements with a high precision. However, the density of PSs is low in non-urban areas and it is imperative to increase the spatial density of measured points. Here, high-resolution SAR sensors offer new opportunities. For this purpose, in addition to the PSs, partially coherent Distributed Scatterers (DSs) can be exploited. Various methods such as the Small Baseline Subset Algorithm (SBAS) and SqueeSAR have been proposed to extract information from DSs. However, SBAS is prone to phase unwrapping errors in rural areas and estimates deformation at only low resolution. The alternative technique SqueeSAR can be computationally expensive as it processes all possible interferogram combinations. Accordingly, this thesis addresses the development of advanced stacking techniques in high resolution SAR interferometry, with a focus on complex areas that are difficult to process using conventional techniques. To this end, first, a new method has been developed for deformation monitoring of DSs at object resolution in non-urban areas. It applies adaptive spatial filtering to improve the differential interferometric phase, followed by deformation estimation using an L1-norm based SBAS approach that is more robust to phase unwrapping errors. Second, an alternative approach for mean deformation velocity mapping of DSs has been proposed for highly decorrelated areas, wherein, wrapped interferograms (with small baselines) are directly used and the deformation velocity is mapped at a suitable object resolution. It includes identification of homogenous patches, estimation of deformation velocity gradients for these patches and then, model-based deformation integration to obtain spatially dense deformation velocity estimates. Lastly, a fusion of TerraSAR-X and TanDEM-X data stacks has been presented for complex urban area monitoring exploiting both PSs and DSs. TerraSAR-X allows monostatic acquisitions and in conjunction with the TanDEM-X satellite, bistatic acquisitions are now also possible. Thus, stacks of monostatic repeat-pass and bistatic single-pass interferograms are available. The bistatic interferograms are of high quality and are free from deformation, atmosphere and temporal decorrelation. By properly integrating the data stacks, an improved estimation of topography and deformation is possible. However, the independent processing and subsequent simple combination of the bistatic and monostatic data is not beneficial. Standard TanDEM-X Digital Elevation Models (DEMs) are inaccurate in dense metropolitan areas because of phase unwrapping errors. These errors occur due to height discontinuities and geometrical limitations such as radar layover. Therefore, the joint processing of TerraSAR-X and TanDEM-X data has been investigated. The developed techniques have been demonstrated using TerraSAR-X/TanDEM-X data from various test sites and a high performance has been proven. The results show an improved utilization of the information hidden in the data and an extension of the applicability of existing techniques for mapping displacement and topography in difficult test areas. The proposed methods can benefit from future SAR systems with even higher resolution.