|Thesis abstract: |
The thesis debates the problem of the external polarimetric calibration and system monitoring of Synthetic Aperture Radar (SAR) devices. More specifically, the efforts has been aimed at assessing the capabilities of natural in-scene scatterers and at proposing novel methodologies for the exploitation of such potential. The first part surveys the calibration feasibility and the distortion estimation performance of the approaches based on distributed targets (DTs). It will be shown that, by introducing convenient assumptions on the distortion model and on the target properties, effective information on the system cross-talks and on the channel imbalance ratio can be extracted from the scene. However, clear evidence that a point calibrator is needed to accomplish a full polarimetric monitoring is also provided. The partial polarimetric calibration achieved by DTs is then investigated with concern to the achievable performance on the well-determinable parameters. A numerical optimization algorithm is proposed to improve the accuracy in case of low channel isolation. The second part is dedicated to an innovative calibration approach based on the stable point targets, namely Permanent Scatterers (PS). The method, hereby called PolPSCal, allows for relative calibration of the full polarimetric distortion matrices (PDMs) affecting the stack images. The algorithm is neither constrained to a particular PDM model nor to any external information, though this latter is needed afterwards to normalize the calculated PDMs. The exploitation of the DT information for the relative calibration of the PolPSCal information is then investigated. The mathematical framework of such over-arching natural target-based approach is reported, and a performance analysis is carried out on a 26 images RADARSAT-2 dataset reporting promising results.