Towards Fully Automatic Generation of Land Cover Maps from Polarimetric Sar Data

Ongoing SAR missions are already providing a large amount of multi-polarization or polarimetric data and more are expected in the next future from new planned missions. Such a scenario suggests fully automatic procedures for mining and effectively exploiting the information embedded in the polarimetric images, especially when multi-temporality is involved. Processing multi-polarization data for classification purposes has been carried out by a variety of supervised algorithms which span from Bayesian Maximum Likelihood to Fuzzy Logic to Support Vector Machines to Multi-Layer Perceptrons (MPL). A training phase carried out under human supervision is usually required by these procedures, thus preventing them from running in a fully automatic mode. To tackle such a crucial issue, the chaining of unsupervised and supervised modules may be considered. In particular, the unsupervised processing stage may rely on the so called Self Organized Maps (SOM) algorithm [1], which produces a self-organizing feature map where specific information is associated with each node position.