[GRASSLIST:6609] surf.rst and segmax, npmin

Hi!
I am trying to interpolate a DEM from an quasi-regular grid of the
region of interest (meaning that I have one data point per resolution
cell, roughly). I am using s.surf.rst (I can't get the data into GRASS
6.0), and I have noticed that if you look at the slope aspect angle
plot of the interpolated surface, the joining lines of the
segmentation procedure are clearly visible in some areas. I am able to
get rid of them by changing the npmin and segmax parameters (say,
segmax=200 and npmin=1000 yields good results). This however, greatly
slows computations, and these values are just trial and error, I don't
really know what set of values will be optimal.

The documentation mentions that if these lines are visible, npmin
needs to be increased, but is there some "recipe" or idea as to the
nature of these numbers?

Thank you very much!
Jose

Hi,

Have You read http://www.ing.unitn.it/~grass/conferences/GRASS2002/proceedings/proceedings/pdfs/Cebecauer_Tomas.pdf
?

I use v.surf.rst (Grass 6.0.0) to create DTMs from isoline contour and
aditional point data. I use tension 35, smoothing 1, segmax 100 and
npmin 400 with good results. Segmentation lineas are visible only at
very undersampled areas. I will fix that by ading some points there
and then will be no problems.
Consider ading some points at areas where segmenation is visible.

WBR,
Maris.

On 4/23/05, Jose Gomez-Dans <jgomezdans@gmail.com> wrote:

Hi!
I am trying to interpolate a DEM from an quasi-regular grid of the
region of interest (meaning that I have one data point per resolution
cell, roughly). I am using s.surf.rst (I can't get the data into GRASS
6.0), and I have noticed that if you look at the slope aspect angle
plot of the interpolated surface, the joining lines of the
segmentation procedure are clearly visible in some areas. I am able to
get rid of them by changing the npmin and segmax parameters (say,
segmax=200 and npmin=1000 yields good results). This however, greatly
slows computations, and these values are just trial and error, I don't
really know what set of values will be optimal.

The documentation mentions that if these lines are visible, npmin
needs to be increased, but is there some "recipe" or idea as to the
nature of these numbers?

Thank you very much!
Jose

Hi Mâris,

http://www.ing.unitn.it/~grass/conferences/GRASS2002/proceedings/proceedings/pdfs/Cebecauer_Tomas.pdf
?

I use v.surf.rst (Grass 6.0.0) to create DTMs from isoline contour and
aditional point data. I use tension 35, smoothing 1, segmax 100 and
npmin 400 with good results. Segmentation lineas are visible only at
very undersampled areas. I will fix that by ading some points there
and then will be no problems.
Consider ading some points at areas where segmenation is visible.

I have read that paper, however the problem is that we have no
accurate topography for the area I'm working in, and thus cannot add
points from other data sources, as we do not know their accuracy (the
differences could be up to 100s of meters in some cases). These errors
would also compromise the accuracy in areas where we have good quality
data.

Any other suggestions? :slight_smile:
Jose