GRASS and Spectral Mixture Analysis

[Markus Neteler]

Hi Simon, hi all!

Maybe I was a little bit unclear - I want to decompose the images
with spectral signatures of different vegetation types, dark and
light soil, water areas etc. You can either take these "endmembers"
of SMA (spectral mixture analysis) of a spectral library or retrieve
it from the image itself by principal component analysis etc. 
The DN value from the satellite image (0..255) is decomposed into 
its summands - so it is a subpixel classification. The results are 
several maps containing the distribution in percentage of the endmembers
you have tested: Some vegetation distribution maps, some soils maps etc.
Others tested for minerals in rather unvegetated areas and so on.

I need a module to perform the required matrix operations with least
square error fitting for each pixel...


> This is definately not the right way but I export it with r.out.tiff and
> then divide the tiff with gimp Image/Channel Ops/Decompose, then I
> r.in.tiff those files and have the three bands of rgb. Copy the cellhd of
> the original tiff to the name of your imports and you'r done.

Simon, this you can do with #-operator in r.mapcalc (see the manpage for
that). From that you can split panchromatic images into RGB or make 
a greyscale equivalent image from color image.
 

Best wishes (and hoping, that someone in GRASS community is working with
SMA)

  Markus Neteler