On Tue, 2010-04-06 at 20:44 +0300, Micha Silver wrote: > joel dinis wrote: > > > Hello Micha, > > > > I put r.info <http://r.info/> -r amostras > > as sugested, get the following: > > min=112 > > max=512 > > > So, if I understand correctly, you are trying to classify into 400 > separate classes?! > I'm a beginner with image classification, and others with more > experience will know better, but I'll bet that's your problem. There > must be many separate training areas (with different cat values) with a > very similar spectral signature, and the algorithm is not able to > differentiate between them.
I 've also thought about that Micha (I was about to suggest Joel to use less training classes), but I was hesitant to say it :-p Nikos > > > (This values are index to my classes.) > > > > JD > > > > On Tue, Apr 6, 2010 at 4:39 PM, Micha Silver <[email protected] > > <mailto:[email protected]>> wrote: > > > > Nikos Alexandris wrote: > > > > (sorry for re-sorting the message, just to keep it easy to > > follow-up) > > > > joel dinis: > > > > > > > I trying to classify remote sensing imagery, and > > > I thought to try the i.smap module, but > > > when I run the i.gensigset, > > > > > > > > Nikos Alexandris: > > > > > > Please, could you copy-paste the exact command you > > used? > > Does the "maxsig=" (e.g. maxsig=20) make any > > difference? > > > > > > > > joel dinis: > > > > > > > > In fact, I put maxsig=30. > > i.gensigset trainingmap=amostras group=grupo subgroup=subgrupo > > signaturefile=assinaturas maxsig=30 > > > > > > > > What about smaller maxsig's, like: > > > > - maxsig=25 > > - maxsig=20 > > - maxsig=15 > > > > ? > > > > Nikos > > > > > > > > Also it might help to know what is contained in the trainingmap. i.e: > > r.info <http://r.info> -r amostras > > > > > > > > > I always > > > get the same problem: > > > ... > > > Removed a singular subsignature number 1 (9 remain) > > > Removed a singular subsignature number 1 (8 remain) > > > Removed a singular subsignature number 1 (7 remain) > > > Removed a singular subsignature number 1 (6 remain) > > > Removed a singular subsignature number 1 (5 remain) > > > Removed a singular subsignature number 1 (4 remain) > > > Removed a singular subsignature number 1 (3 remain) > > > Removed a singular subsignature number 1 (2 remain) > > > Removed a singular subsignature number 1 (1 remain) > > > Unreliable clustering. Try a smaller initial > > number of > > clusters > > > Removed a singular subsignature number 1 (-1 remain) > > > Unreliable clustering. Try a smaller initial > > number of > > clusters > > > Number of subclasses is 0 > > > Clustering class 12 (184 pixels)... > > > Removed a singular subsignature number 1 (9 remain) > > > Removed a singular subsignature number 1 (8 remain) > > > Removed a singular subsignature number 1 (7 remain) > > > Removed a singular subsignature number 1 (6 remain) > > > Removed a singular subsignature number 1 (5 remain) > > > Removed a singular subsignature number 1 (4 remain) > > > Removed a singular subsignature number 1 (3 remain) > > > Removed a singular subsignature number 1 (2 remain) > > > Removed a singular subsignature number 1 (1 remain) > > > Unreliable clustering. Try a smaller initial > > number of > > clusters > > > Removed a singular subsignature number 1 (-1 remain) > > > Unreliable clustering. Try a smaller initial > > number of > > clusters > > > Number of subclasses is 0 > > > i.gensigset complete. > > > > > > And afterwards the signature file doesn't have any > > > spectral information. I really don't understand whats > > happening > > > and the cause to this outcome. > > > > > > Can anyone solve this "enigma"? > > > > > > Ps: I don't know if this information is important: I > > collected > > > about 450 samples distributed by 10 classes. I > > may also > > > add that the average number of pixels per sample > > is around > > 50. > > > > > > And thanks list, for the previous tips: they were > > very > > useful. > > > > > > > > > > > > _______________________________________________ > > grass-user mailing list > > [email protected] <mailto:[email protected]> > > http://lists.osgeo.org/mailman/listinfo/grass-user > > > > This mail was received via Mail-SeCure System. > > > > > > > > > > > > > > -- > > Micha Silver > > Arava Development Co. +972-52-3665918 > > http://www.surfaces.co.il > > > > > > > > > > This mail was received via Mail-SeCure System. > > _______________________________________________ grass-user mailing list [email protected] http://lists.osgeo.org/mailman/listinfo/grass-user
