[GRASSLIST:4876] image classification question

[eicksch]

Hi all,

I have two sets of image/raster data of the same region acquired at
different times. I am able to perform unsupervised radiometric
classifications on each of them. Just for interest I would like to classify
each of the scenes using the signatures defined for the other image (group).
For instance, to classify scene2 with the signatures from scene1, I cd'ed to
        $LOCATION/group/group2/subgroup/subgroup2/sig
and made a
        ln -s $LOCATION/group/group1/subgroup/subgroup1/sig/sigfile1 .
Unfotunatly,
        i.maxlik group=scene2 subgroup=subscene2 sigfile=sigfile1 
class=scene2.sig1
results in
        Can't read signature file sigfile1
I fear, that it isn't related to the existance of the file but all the
contents (the number of pixels per class, the variance matrices etc.).
What did I make wrong and how can I get my results?

In other words: How can I generate a (valid) signature file by hand?

All this applies to GRASS 4.2.1.

Thanks in advance for any help,

Stephan

--
------------------------------------------------------------------------
Stephan Eickschen                          Tel. : +49 (0)8153 28-1351
GeoForschungsZentrum Potsdam (GFZ)         Fax. : +49 (0)8153 28-1207
Division 1, Section 1.2
c/o                                        email: eicksch at gfz-potsdam.de
DLR Oberpfaffenhofen
82234 Wessling
Germany
------------------------------------------------------------------------