We present a method to determine spectral differences and compositiona
l variability of planetary surfaces when remotely sensed by means of i
maging spectroscopy instrumentation. The quantity frequently measured
in remote sensing of planetary bodies in the 0.3-2.5 mu m spectral ran
ge is the reflectance spectrum and it is used to study the mineralogic
composition of the surface being sensed. Very often, however, this qu
antity is difficult to measure owing to lack of knowledge either of at
mospheric extinction or of analogous ar spectra. The method we describ
e allows all the in a surface having spectral similarity, to be identi
fied and does not require the reflectance spectrum measured, taking fu
ll advantage of imaging spectroscopy and classification methods. In pr
actice, all the sco spectra of the image are divided by the spectrum o
f a pixel (or an average spectrum) tak-en in the image itself. Ire Swa
y, both the instrument and the atmospheric transfer functions are elim
inated and it is possible to compare the spectra with one another. The
se new relative spectra are then classified using a clustering algorit
him in order to recognize spectral similarities. The final result is a
map showing all the regions in the image having similar spectra which
can be linked to the mineralogical composition of the surface under T
he method has been applied to Earth-based observations of the Moon but
can be equally used with; other high spatial imaging spectroscopy dat
a provided by future interplanetary missions. (C) 1997 Published by El
sevier Science Ltd. All rights reserved.