[GRASS5] image analysis question i.clust, i.maxlik, i.smap

Michael Barton michael.barton at asu.edu
Mon Apr 26 18:50:40 EDT 2004


This is a question for those of you with experience in classification 
and feature extraction from multiband imagery.

I've been using MultiSpec (a program I can highly recommend, that is 
also public domain) and a couple other pieces of software, but want to 
start using GRASS for at least some work. So, I want to do some 
comparisons of GRASS's supervised and unsupervised classification 
routines with those of other programs I use.

I have a 14-band ASTER image I've imported into GRASS and want to start 
with an unsupervised classification in a search for a particular set of 
prehistoric earthworks.

I set up the required groups and subgroups. Then I ran i.clust and it 
appeared to work fine. I started with 25 classes, as I am looking for 
some rather fine-scale features, and left the rest of the defaults 
alone for now. The output text said it ran the appropriate iterations. 
I had it save a text report and this looked OK. The signature file 
*looks* OK also and was placed by GRASS in the 
/group/[name]/subgroup/[name]/sig folder.

I then tried i.maxlik. It produced an output to screen of 25 lines that 
said

"WARNING: signature 1 is not valid (ill-conditioned) - ignored"
and so forth for signatures 1-25, followed by
i.maxlik ... 0%

It produced no classification map obviously.

I also tried the the i.smap classification routine using the same 
signature file. It also produced an error:
"WARNING: unable to open signature file AST_1336_clustsig for subgroup 
nonthermal_bands of group [AST.1336 in satellite].
ERROR: signature file [AST_1336_clustsig] missing or unreadable."

I haven't misspelled anything. I also output my GRASS data (same region 
window and everything) to a *.lan file and read it into Multispec. I 
had no trouble doing an unsupervised classification and (happily) it 
produced potentially interesting results without even futzing with the 
default values.

Any suggestions?


______________________________
Michael Barton, Professor & Curator
Department of Anthropology
Arizona State University
Tempe, AZ  85287-2402
USA

voice: 480-965-6262; fax: 480-965-7671
www: http://www.public.asu.edu/~cmbarton
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: text/enriched
Size: 2203 bytes
Desc: not available
Url : http://lists.osgeo.org/pipermail/grass-dev/attachments/20040426/c24a2463/attachment.bin


More information about the grass-dev mailing list