Dear list,
I know this topic has already been discussed at times, but I couldn't find any useful answers on the net.
I'm trying to do some supervised classification with. i.gensig and i.maxlik.
Unfortunately, i,gensig creates signature files without the "category 1" ... "category n" labels, so i.maxlik creates a classified raster file with wrong class numbers assigned to pixels.
Therefore, if categories 1, 3, 5 and 11 are represented in the original raster (the one that must be classified), they will be represented as 1, 2, 3, 4 in the classified raster.
How do I get the category labels in the signature files?
Thanks.
Paolo
On 18/05/12 08:21, e-paul wrote:
Dear list,
I know this topic has already been discussed at times, but I couldn't
find any useful answers on the net.
I'm trying to do some supervised classification with. i.gensig and
i.maxlik.
Unfortunately, i,gensig creates signature files without the "category 1"
... "category n" labels, so i.maxlik creates a classified raster file
with wrong class numbers assigned to pixels.
Therefore, if categories 1, 3, 5 and 11 are represented in the original
raster (the one that must be classified), they will be represented as 1,
2, 3, 4 in the classified raster.
You can use r.reclass.
Moritz
How do I get the category labels in the signature files?
There are no category labels in the signature files. You have to make the match visually.
i.gensigset + i.smap gives you the "right" category values for supervised classification results, i.e. values based on those used in the training set. It should be possible to reuse the code from i.gensigset and i.smap to do the same for i.gensig + i.maxlik.
You can file an enhancement request in the bug tracker.
Moritz