On 19/06/18 19:32, Jamille Haarloo wrote:
Hi Moritz,
Thanx for the advice! Will check it out. From the publications I read, I
inferred it was better to use an alternative method (data mining/
object-based) on high resolution imagery than pixel-based.
Generally this is true, but when you work in landscapes which don't
really have sharp edges, this can sometimes be counterproductive. There
is not clear scientific answer, though, and you just have to try what
works best.
I also read
that texture metrics (e.g. standard deviation of spectral bands) may
capture the unique “patterns” of wetlands and other ecosystems (given
the appropriate spatial scale). Will continue testing and checking for
improvements.
I think in this case "real" texture measurements (e.g. those provide by
r.texture, might even be more efficient than just the standard
deviations of bands. Again, you will have to do a bit of trial and error
to determine which texture measure works best.
I was running directy from GRASS GIS, but recently created and used a
new training map with the lower case column name. My apologies for the
inconsistency, but good to know that the previous training map wasn't
the issue.
Actually I noticed that your error message concerned the qbwwv voting
algorithm which wasn't in the command line you provided. I was just able
to reproduce your error using that same voting scheme. I'll have to dig
into this a bit deeper to see where the error comes from.
Moritz
_______________________________________________
grass-user mailing list
[email protected]
https://lists.osgeo.org/mailman/listinfo/grass-user
