Beau Bush (bbush@ticsys.tamu.edu) writes on 27 Jul 94:
Could someone tell me if GRASS has the capability to type vegetation areas
using infrared reflectance. I have a key made from a small portion of the area that we are studying that has each vegetation type catagorized. I have been
told that this is possible but I have no idea how to do it. This was not my
idea and I did not start this task but it has been left for me to finish. Is
there software that will do such a thing? Any correspondance would be
appreciated. If I need to elaborate on this please let me know.
[my apologies if this sounds rudamentory. Hopefully this will be
of help to a few.]
Do you have several reflectance images, each representing a different
wavelength? If so, you want to do something called "supervised
multispectral classification."
There are two separate algorithms in GRASS for doing this:
maximum likelihood and sequential maximum a prior estimation.
(i.gensig/i.maxlik and i.gensigset/i.smap)
i.gensig/i.gensigset work on your training data. They statistically
define where in n-dimensional space lies each type of vegetation.
i.gensigset can generate "multimodal" training statistics
(a strength of i.smap)
i.maxlik/i.smap do the classification, which is the next step.
These two steps are fairly easy. It's the preparation that might
make things confusing to some. Particularly, importing the
data and somehow figuring out how to assign/delineate training data.
For your case, perhaps r.in.poly might be the best route since
you already know which sections of the image to do the training on.
Also, 'g.manual imagery' will help you understand the GRASS concept
of "groups," which is important for every step of the process.
[so, which do you use, maxlik or smap? depends on your "scene"
and your objective. there's a draft manuscript at
ftp://pasture.ecn.purdue.edu/pub/mccauley/papers/smap-echo-ml-compare.ps.gz
where Bernie Engel and I look at the performance of i.smap versus
i.maxlik and another algorithm. i.smap did the best for "homogeneous"
fields (usually found in cultivated scenes).]
Beau Bush
bbush@ticsys.tamu.edu
--
Gig 'em,
Darrell McCauley '90 '92
P.S. (for Aggies only) Robert Maggio over in Forest Science may be
a good person to get to know if you're going to be doing a lot
of remote sensing work.