Hi all,
Ok, I successfully could do a part of what I want.
I had to first export the classes into ROIs, then import the new ROIs
with the Endmember Collection tool, and finally (that's what was
missing) change the "input file" to the whole area. I was then able to
use some algoriths (no max. likelihood allowed however!).
This is a problem. The resulting classification is quite different on
the subset because the first classification was done unsupervised
(Isodata) while the second one was supervised with the ROIs (e.g. SAM,
Minimum Distance).
Anyone has a clue how to get the -same- classification results while
going from an unsupervised to a supervised classification?
Thanks again :-)
Alex
Alexandre Leroux wrote:
> I did an unsupervised classification over the subset of an area. I want
> reuse the same classes (as 'endmembers') to classify the whole area.
>
> If I try any supervised classification, I am asked to provide ROIs
> instead of classes. Even if I export my initial classes from the
> subsetted area into ROIs (long process), they cannot be used on the
> whole area (different extent?).
-- Alexandre Leroux, M.Sc., Ing.Sta. Environnement Canada Division des Urgences Environnementales alexandre.leroux@ec.gc.ca
This archive was generated by hypermail 2b29 : Tue Nov 09 2004 - 11:56:33 PST