Discrimination of land cover from a multiparameter SAR data set

The identification of the most valuable radar observation parameters (e.g., frequency, polarisation, incidence angle) is important both for designing nonredundant high-performance sensors (i.e. selection of frequency bands an d polarisations) and for specifying mission operation requirements (i.e. te mporal sampling, incidence angle). Moreover, the task of classifying multip arameter SAR images may require to adopt a strategy that implies the select ion of a number of features among those available from this kind of sensors . In this paper we have performed this kind of analysis in a specific area of interest to account for the particular conditions in which remotely sens ed data are going to be used. The paper summarises the results of the analy sis of the radar data acquired during the MAC Europe '91 and X-SAR/SIR-C ca mpaigns over the Montespertoli test site in Italy. The analysis is based ma inly on a statistical approach aiming at demonstrating what is the contribu tion of different measurements performed by the polarimetric SAR for discri minating the surface coverage. The work is intended to furnish a guideline to develop an optimal strategy for acquiring and processing polarimetric da ta to be used for land classification.