Therefore, MLC requires a preliminary
knowledge necessary to generate representative parameters for each
class of interest and to carry out the training stage. The whole procedure,
as for all the supervised classification algorithms, requires two
basic steps: (i) clustering or training which consists in providing
known areas for each class, generally identified through in situ analysis,
and (ii) classificationwhich is carried out by comparing the spectral signature
to each pixel (under investigation)with the spectral signature of
the training cluster.