[GRASS-user] supervised classification "no subgroup files available"

Silvia Simoni silvia.simoni at ing.unitn.it
Fri Sep 5 04:28:37 EDT 2008


Hi Markus,

As to the training area I draw just one, which is fearly small because  
it's homogeous with rispect to the color I want to sample. I've tried  
to select other areas in the same section, using the command "draw  
region", but apparently I'm not allowed to do it.

I always close the training areas with the "complete button".

Here it's the procedure I'm following
1. i.group (to create group A and subgroup rgb)
2. d.mon
3. i.class map=ortophoto_composite group=A subgroup=rgb outsig=test (I  
can see the file test being created in the folder .../subgroup/ 
$SUBGROUP/sig/; it's now 0 Byte)
4. from the interactive i.class menu, I select "Zoom"; then once the  
zoomed area is displayed, "Define Region"  and "Draw Region". I draw 1  
area by points, then I click on "Complete Region" and "Done". As I  
return to the main menu, I click on "analyse region" -> Signature  
Menu, "Display matches", "Done". save Signature YES, then I'm prompted  
for the signature description; finally I Quit from the main menu. Now  
I can see the file size has increased to 14 Byte. When I edit it, I  
can see only 3 lines.

I've also tried selecting a much larger area for which the histogramm  
is smoother that in the previous case, nothing has changed, still 3  
lines in the outsig file.

Thanks,

Silvia

On Sep 4, 2008, at 9:52 PM, Markus Neteler wrote:

> On Thu, Sep 4, 2008 at 6:24 PM, Silvia Simoni
> <silvia.simoni at ing.unitn.it> wrote:
>>
>> Thanks Markus!
>>
>> I'm now using grass6.4.svn but I still have trouble with the  
>> supervised
>> classification process. Now I can run i.class and produce the  
>> spectral
>> signature for my map but this file is NOT readable by i.maxlik. I  
>> get the
>> following message:  Can't read signature file [signaturefile].
>>
>> I checked in $MAPSET/group/subgroup/$SUBGROUP/sig/; the  
>> signaturefile is
>> there but it only contains 3 lines, as follows:
>> #
>> #comment
>> single value XXXX
>>
>> Is this correct? I obtained other signature files from i.cluster  
>> which look
>> different from this. It seems to me that i.class doesn't write the  
>> signature
>> file correctly so that i.maxlik is not able to read it.
>
> It doesn't look good.
> But a quick test here didn't show any problem.
> Are you sure that
> - your training areas are big enough in i.class?
> - that you close the training areas with the "complete" button?
>
> Markus

----------------------------------------------------------------------------------------
Silvia Simoni PhD
DICA - Department of Civil and Environmental Engineering
Universita' di Trento
via Mesiano 77, 38100 Trento, ITALY
tel. +39 0461 882610
fax. +39 0461 882672
e-mail: silvia.simoni at ing.unitn.it





More information about the grass-user mailing list