Temporal Monitoring of Polarimetric SAR Systems by Natural Calibrators
Iannini, Lorenzo¹; Monti Guarnieri, Andrea²; Tebaldini, Stefano^2
1TU-Delft, NETHERLANDS; ²Politecnico di Milano, ITALY

So far, the calibration of full-polarimetric SAR systems mostly relies on ground installations of calibrated active (transponders) or passive (corner reflectors) targets. It is however evident that part of the system health information can also be retrieved from other natural features which are intrinsic to the illuminated scene. The main research question is then how the scene information must be collected and processed in order to produce an effective contribution to the assessment of the PolarimetricDistortion Matrices (PDMs). Two main conceptual categories of in-scene information processing can be identified:
1. the model extraction from the scatterer spatial statistics
2. retrieval of coherent information from temporal target observation series
The first one exploits distributed target (DT) areas and has already been widely explored in literature. Without a calibrator, the DT-based approach provides unambiguous and reliable estimates of the cross-talks of the imbalance ratio, though they cannot guarantee radiometric and phase consistency of the channels throughout the acquisitions. The second approach represents the most innovative contribution of this work: the exploitation of Permanent Scatterers (PSs) in polarimetric calibration activities. The novelty in the solution proposed required efforts in the formalization of an appropriate PS framework. The fundamental steps of such process were the transposition of the traditional PS model to the polarimetric context and the development of an effective methodologyfor robustly tackling the inverse problem. This led to the definition of a polarimetric PS detection based on the Generalized Likelihood Ratio Test criterion and to an efficient estimation procedure which exploits the low rank nature of the phase-calibrated PS stack. The integration of these steps in a joint iterative algorithm was named PolPSCal. It will be shown that the theoretical calibration performance is in a first approximation related to the number of PSs detected in the imagette and to their polarimetric quality and that it is indeed independent from the PDM expression. The algorithm is neither constrained to a particular PDM model nor to any external information, thus its implementation is practically feasiblefor any SAR sensor. The returned PDMs contain all the required distortion information, though in an ambiguous way. In order to remove part of the ambiguity when no calibrated images are available (a single one would allow indeed a full disambiguation of the whole stack) the estimatesfrom the DT are employed. It will be shown that thetwo information (PS and DT) are in fact complementary and thus their joint combination can provide an effectivecontribution to polarimetric monitoring. Once again, a generic distortion model hasbeen conceived, so that the approach can be readily tailored to different sensor scenarios,ranging from higher-frequency SARs to lower-frequency acquisitions affected by Faraday rotations. When applied to C-Band the encompassing framework achieves anunambiguous temporal monitoring of the all the distortion parameters, up to an absolute radiometric scale factor. The performance has been assessed through synthetic simulations and validated on a Fine Quad-Pol Radarsat-2 dataset which comprises 26 images over theBarcelona area, returning an accuracy on the channel imbalance stability below 0.2 dBand on the cross-talk level <-35 dB, some promising results indeed, that comply with the acknowledged quality standards of polarimetric SAR systems.