[GRASSLIST:2881] Re: visible segments in the DEM
neteler at itc.it
Thu Jan 10 11:37:56 EST 2002
A forward from Helena Mitasova:
On Thu, Jan 10, 2002 at 11:01:13AM -0500, Helena Mitasova wrote:
> I do not subscribe to grassusers list.
> Quick answer is - use s.surf.rst, substantially reduce the number of
> points on contours - they are oversampled (choose dmin at least 1.5 times
> resolution and dmax a BIG number so that the program does not add
> additional points to contours). If you still get the segments - run
> r.random on your new DEM in the area where you have the visible segments
> to generate few additional points and add them to your contours and
> reinterpolate. And use smoothing at least 0.1 (you can use variable
> smoothing with s.surf.rst - set it zero for the peaks and higher value for
> the flat areas).
> I am thinking about taking out v.surf.rst from the distribution as it
> needs all kinds of fixes (it gives seg fault if you compute the curvatures
> - there is a bug in writing the color tables, but it computes everything
> OK, also the default tension is too high, and overall I am not sure what
> it is doing - what was updated and what not - e.g. how could it run with
> npmin=2000 - it has a 400 or 700 limit).
> Also I have the fix for s.surf.rst for preventing it to run forever when
> it gets dense points (such as LIDAR), just 2 lines of code, but I am
> hesitating to include it for release - I would need testers to run it with
> various types of data to see whether the fix does not have uninteded
> consequences (I will forward the fix in the next email)
> It is not too difficult to fix the visible segments problem - the errors
> is very small so it is more cosmetics (but a very unpleasant one) - so
> those segments just need to be smoothed out withing the program - we used
> to have it with older segmentation that we used for the MG93 paper).
> If anybody with good programming skills would be interested in making
> improvements and new developments for multivariate interpolation for GRASS
> part of their research (even PhD) I will be very happy to work with
> him/her. I have a long list of improvements that need impmenetation,
> testing enhancement etc.
More information about the grass-user