Advances and Experiments of Tomographic SAR Imaging for the Analysis of Complex Scenarios

It is expected that the number of synthetic aperture radar (SAR) images available for a same scene will increase exponentially in the future, thanks to the technical developments in this area. In order to fully exploit the information lying in data acquired in looking angle (multibaseline, MB), time, and polarization diversity, developments are underway of processing techniques which constitute an evolution of the mature phase-only SAR interferometry for producing new and/or more accurate measures. In particular, by combining coherently (i.e. amplitude and phase) the SAR data, new opportunities are arising for an improved imaging and information extraction of the observed scene. Among these techniques, a very promising advance is constituted by SAR Tomography, a MB interferometric mode allowing a full 3-D imaging in the range-azimuth-height space, thus separating multiple scatterers in layover at different heights in the same SAR cell in complex scenarios. Recently, a new interferometric mode called Differential SAR Tomography has been conceived at the University of Pisa from the synergic fusion of SAR Tomography and the conventional Differential Interferometry, allowing the estimation of also the possible relative deformations between multiple layover scatterers. In this thesis, theoretical advances and experimental results are presented in the analysis of complex scenarios. In particular, the tomographic imaging problem is addressed by exploring different algorithmic options able to enhance the image contrast and possibly also increase the scatterer resolution in height. Moreover, in order to automate the estimation of the height or height/deformation velocity, a scatterer detection algorithm has been developed, which constitutes also a preliminary step for the extensive validation of the information extracted. With regards to volumetric scatterers (e.g. the scatterer in forest scenarios), tomography-based coherent data combination techniques have been proposed and investigated, in particular for the extraction of the sub-canopy digital terrain model and for deriving in a non-model based fashion a coherent MB dataset with only the signal from the scattering layer of interest. Finally, the differential tomographic framework has been exploited for the robust tomographic analysis of temporal decorrelating volumetric scatterers. For each investigated topic, extensive experiments have been carried out with MB urban and forest SAR data.