Image classification

[IBH7CAB at MVS.OAC.UCLA.EDU]

You can rectify first, if spatial interpolation is done by
nearest neighbor resampling, not a polynomial convolution.  Perhaps
the best way to work with GRASS is to (1) rectify a color composite
of your satellite image, (2) classify your unrectified bands, then
(3) apply the transformation coefficients estimated in step 1 to
rectify the classified image.