Chapter
6
Conclusions and Perspectives
In this thesis, tomographic and differential tomographic techniques have been proposed for the coherent processing of multibaseline-multitemporal SAR data collected over complex (possibly non-stationary) scenarios. Extensive experiments have been carried out with urban and forest data acquired by spaceborne and airborne radar platforms.
The ABF processing has been experimented for high contrast 3-D tomographic imaging, demonstrating a superior performance w.r.t. the conventional Fourier-based imaging. ABF is radiometrically non-linear, and can lead to height super-resolution. However, to work properly it needs a multilook averaging, thus reducing the horizontal resolution of the to-mographic reconstruction. An interesting future advancement can be the extension of the adaptive processing also to handle single look data, along the lines of the first experiments in [83]. For all the applications in which the radiometric fidelity is an important requisite (e.g. for biomass studies), a high contrast linear alternative to ABF is the knowledge-based multibaseline interpolator, which makes use of a light a priori information about the height sector containing the scatterers. In this case, sidelobe reduction is possible also with single look data, however no super-resolution is provided. A way to be investigated for obtaining also some resolution enhancement could be the extrapolation of the data samples to baselines greater than those available.
Independently of the imaging method, the common band data filtering has been demon-strated to be an effective pre-processing step to mitigate the perspective effects along the height dimension introduced by low bandwidth radar signals, which is a strict requirement for P-band data acquisitions in order to avoid electromagnetic interferences with ground emitting systems. Requirements on the minimum signal bandwidth have been discussed for the applicability of the common band pre-filtering without loss of information in dependence of the looking geometry. However, it has been noticed that the minimum bandwidth could be sufficiently high to not introduce noticeable perspective effects, or even higher than the band-width constraint. Future research could be devoted to investigate alternative solutions to the common band pre-filtering in order to reduce the perspective effects without independently from the looking geometry. An interesting and original technique worth to be investigated could be the application of a 2-D high resolution focusing in the range-height plane.
Tomo-SAR has also shown to be a powerful tool for other applications involving the coherent combination of SAR data. In particular, by means of experiments with P-band data,
Chapter 6 - Conclusions and Perspectives
parametric and not Tomo-SAR based techniques have been shown to achieve accuracies in the order of magnitude of 1 m in the extraction of the ground heights in forested scenarios. Other investigations could be fruitfully carried out to assess the ultimate height estimation accuracy obtainable with forest data collected in L-band, in which the canopy contribution becomes non-negligible and may impair the ground height extraction if specific procedures are not set up. In this case, especially if the height resolution is a critical factor due to the non-optimal baseline aperture, model-based estimation techniques, possibly based on low computational burden maximum likelihood estimators, could be employed to enhance the accuracy and to provide a richer output for the description of the observed scenario.
To automate the scatterer height and height/deformation velocity information extraction in the Tomo-SAR and Diff-Tomo frameworks, respectively, a sub-optimal scatterer detection algorithm has been proposed combining ABF and a model-based fitting in the complex data domain. The experiments carried out with urban spaceborne data have demonstrated the ef-fectiveness of the algorithm, and the possibility to extract information of layover scatterers with a sub-Rayleigh height distance thanks to the adaptive processing. The scatterer detec-tion has also been used as a preliminary validadetec-tion of the height estimated through ABF by geocoding the coordinates of the detected scatterers. First experiments have also been carried out by reducing the multilook degree of ABF. Research efforts could be spent in the devel-opment of single look detectors. A possibility could be the implementation of the hybrid scatterer detection using the single look adaptive processor of [83]. Alternativaly, also the sector interpolator could be a good candidate for getting scatterer identification at the highest horizontal resolution, also employed in an iterative fashion as in [79]. It is expected a re-duced scatterer mislocation and/or false alarm probability with respect to the linear inversion framework of [21] given the augmented sidelobe suppression thanks to the interpolation.
Finally, the capabilities of the Diff-Tomo processing in carrying out tomographic analy-ses of temporal decorrelating volumetric scatterers have been shown, for getting non-blurred and resolution-restored Tomo-SAR profiles and enhancing the height estimation accuracy with respect to a Tomo-SAR only processing. In this case, future work could regard the improvement of the estimation efficiency e.g. by resorting to low complexity maximum likelihood-based estimators.
It is worth remarking that part of Tomo-SAR and Diff-tomo techniques presented here have been investigated in the framework of the IP FP6 LIMES (Land and sea Integrated Monitoring for European Security) project, GMES (Global Monitoring for Environment and Security) Program, funded by the European Commission. Moreover, studies for new space-borne missions may benefit from application of Tomo-SAR and Diff-tomo concepts, such as the ESA candidate Earth Explorer Core Mission BIOMASS for global biomass monitor-ing, investigated in the series of the BioSAR projects. Another example may be the DLR TanDEM-L mission for the observation of the Earth system and its climate dynamics, where the multistatic acquisition scenario will be applied. The multistatic investigations may be also fruitfully applied to the incoming DLR TanDEM-X system, also concerning the urban subsidence monitoring, and to possible advanced orbital configurations of COSMO-SkyMed.
