[GRASS-user] multi-date image segmentation in grass

Hi,

I have several Landsat (4-7) images from 1990-2002 which will be used
for change analysis of land cover. FAO's FRA-RSS [1] developed a
standard methodology for integrating multi-temporal Landsat into a
consistent sample tiles. The article [2] outlined the general method
as:
- Image subsetting
- Radiometric normalization
- Contrast enhancement
- Multi-date image segmentation

Most of the processing procedures I was able to do in GRASS GIS,
except for the last step. The process combined the multi-date imagery
where an automated image segmentation was implemented to derive
preliminary polygons to identify changes. I believe they used
ecognition for the segmentation. I've used i.smap for preliminary
segmentation on separate dates but not for a combined multi-date
imagery. Anyone tried this with GRASS? Can i.pca do this or another
tool?

Any idea

[1] http://www.fao.org/forestry/fra/remotesensingsurvey/en/
[2] http://www.informaworld.com/smpp/content~content=a923063789~db=all~jumptype=rss

--
cheers,
maning
------------------------------------------------------
"Freedom is still the most radical idea of all" -N.Branden
wiki: http://esambale.wikispaces.com/
blog: http://epsg4253.wordpress.com/
------------------------------------------------------

On Tue, Sep 28, 2010 at 1:12 PM, maning sambale
<emmanuel.sambale@gmail.com> wrote:

Hi,

I have several Landsat (4-7) images from 1990-2002 which will be used
for change analysis of land cover. FAO's FRA-RSS [1] developed a
standard methodology for integrating multi-temporal Landsat into a
consistent sample tiles. The article [2] outlined the general method
as:
- Image subsetting
- Radiometric normalization
- Contrast enhancement
- Multi-date image segmentation

Most of the processing procedures I was able to do in GRASS GIS,
except for the last step. The process combined the multi-date imagery
where an automated image segmentation was implemented to derive
preliminary polygons to identify changes. I believe they used
ecognition for the segmentation.

We can (start to) recommend r.seg from Addons for this (I will ask
the author to accept inclusion in GRASS 7 as i.segmentation). We
could greatly improve the performance of i.smap by preprocessing
data with r.seg.

I've used i.smap for preliminary
segmentation on separate dates but not for a combined multi-date
imagery. Anyone tried this with GRASS? Can i.pca do this or another
tool?

I didn't try but it sounds interesting also for us (just lack of time). If could
provide me with a small spatial excerpt of all the channels as location,
I would try a few ideas and report.

Markus

Any idea

[1] http://www.fao.org/forestry/fra/remotesensingsurvey/en/
[2] http://www.informaworld.com/smpp/content~content=a923063789~db=all~jumptype=rss

--
cheers,
maning
------------------------------------------------------
"Freedom is still the most radical idea of all" -N.Branden
wiki: http://esambale.wikispaces.com/
blog: http://epsg4253.wordpress.com/
------------------------------------------------------
_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user