Fuzzy contextual classification of multisource remote sensing images

Our objective has been to model satellite image classification
as a cognitive process, providing a procedure that mimics
the rich interaction of human activity in solving classification
problems. The key features of this approach are the definition of a
knowledge-based classification methodology designed to integrate
contextual information into a multisource classification scheme,
together with a fuzzy knowledge representation framework to
model the overall process in a form that closely resembles the
mental representation of human experts. An application for the
identification of the glacier equilibrium line in two different
zones of the Italian Alps has been developed to evaluate the
performance of our methodology in a real domain where class
discrimination requires the simultaneous use of contextual and
multisource information. Numerical results are provided and
compared with those obtained by a conventional classification
procedure. The advantages of the approach, as seen in the
experimental context, are examined.