Geographic Information Systems Stack Exchange is a question and answer site for cartographers, geographers and GIS professionals. It only takes a minute to sign up.
Sign up to join this community
I have a b&w image i want to classify; the trouble is that mere spectral classification yields a lot of error, I saw this in wikigrass
In case of panchromatic maps or limited amount of channels, it is often recommended to generate synthetic channels through texture analysis (r.texture)
And there's nothing more about it in that page; How can I use the outcome from r.texture in a classification, together with the segments of i.segment or i.smap?
After researching a lot, I've found/developed this GEOBIA workflow for GRASS GIS in combination with R machine learning:
in GRASS: calculate the texture with r.texture; make the objects with i.segment and vectorize this output with r.to.vect, then extract the raster's graylevel and texture features with v.rast.stats then choose some training objects and write a shapefile and a csv with v.out.ogr
in R, with RWeka: divide your csv or dbf in training and test sets, apply J48() and then predict(), bind the predicted vector to a column in your data.frame and save the dbf, then you'll have a fitted column
