All,
I sent this to Markus and he suggested that some folks on the list might
be interested as well.
A recent article in the International Journal of Remote Sensing
described a comparison of image classification results from the random
forest and kNN algorithms using GRASS and R. The title is "A comparative
analysis of kNN and decision tree methods for the Irish National Forest
Inventory" and it is in Vol 30 No. 19 October 2009 pages 4937-4955. The
authors are: Daniel O. McInerney and Maarten Nieuwenhuis.
A while back I mentioned I was doing some image classification work
using random forest using R and GRASS. I ended up not using the GARSS-R
interface since my images were quite large (30 GB) so I do pretty much
everything (training area digitizing is done in QGIS) in R. I am working
on a tutorial for using R to do image classification and another for
creating percent cover maps. The percent cover tutorial uses a GRASS
script to create a table of reference samples the regression analysis in
random forest.
I am using GRASS more and more and am encouraging other folks to do the
same. We have a small grant in SE Asia to establish small training and
data centers and I hope to use that as an opportunity to develop more
outreach and training stuff.
All the best,
Ned
_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user
