Experimental and model investigation on radar classification capability

The capability of multifrequency polarimetric synthetic aperture radar (SAR ) to discriminate among nine vegetation classes is shown using both experim ental data and model simulations. The experimental data were collected by t he multifrequency polarimetric AIRSAR at the Dutch Flevoland site and the I talian Montespertoli site. Simulations are carried out using an electromagn etic model, developed at Tor Vergata University, Rome, Italy, which compute s microwave vegetation scattering. The classes have been defined on the bas is of geometrical differences among vegetation species, leading to differen t polarimetric signatures. It is demonstrated that, for each class, there a re some combinations of frequencies and polarizations producing a significa nt separability. On the basis of this background, a simple, hierarchical pa rallelepiped algorithm is proposed.