MULTIVARIATE CLASSIFICATION OF MULTISPECTRAL IMAGES - AN APPLICATION TO ISM MARTIAN SPECTRA

The powerful statistical multifactorial method Gmode is described and its application to the problem of the classification of multispectral imaging is discussed. An application to the classification of Martian multispectral data obtained by the imaging spectrometer ISM on board t he spacecraft PHOBOS2 is presented; a test region is analyzed in this paper, the region of Pavonis Mons. The application of Gmode to the cla ssification of multispectral images has been found to be very promisin g. Automatic mapping of mineralogically different provinces can be per formed using our classification procedure; also, it is possible to sea rch for a specific spectral signature within an otherwise homogeneous region. The most important factors affecting ISM spectra were automati cally identified by the Gmode classification as follows: 1) ''Albedo'' effects; 2) Atmospheric effects; 3) Slope and overall shape of the sp ectra.