[GRASS-user] detection of landscape features from landsat

[Markus Neteler]

On Sat, Jan 24, 2009 at 3:52 AM, Kenneth Brookes
<ken_brookes at hotmail.com> wrote:
>
> Hi Timmie
>
> I am not very familiar with GRASS GIS either but I'm trying to learn.
> However, GRASS does have the tools to perform multi-spectral image
> classification.

Besides that you can also generate indices or do a transform.
See for example
 http://grass.osgeo.org/grass64/manuals/html64_user/i.tasscap.html
 -> 'Wetness' Tasseled Cap component 3 (screenshot)

which identified the lakes near Raleigh, North Carolina. Thresholding
that image will extract the lakes.
Using additional water-sensitive indices and image classification
(perhaps also r.texture to generate synthetic channels) will help
to create a set of potential water body maps.
Finally, adding them with r.mapcalc gives a kind of probability map
for water bodies. The result can then be thresholded (r.mapcalc)
to generate a Yes/No (1/null) map ready for vectorization with
r.to.vect.

Just some ideas,
Markus