[GRASS-user] LiDAR interpolation (v.surf.bspline)

[Markus Metz]

Carlos Grohmann wrote:
>
> Now, I'm working with LiDAR data. From the wiki, interpolation with
> v.surf.bspline should be used, and:
>
>> # Estimation of lambda_i parameter with cross validation (watch for RMS!)
>> # and note use of bicubic for DSM and bilinear for DTM here and below
>
> My question is *why* use bicubic for DSM and bilinear for DTM?
>
> Is that because a bilinear will result in more sharper limits for buildings?
>
I think DSM (Digital Surface Model) means buildings and trees should
be present in the DSM, whereas a DTM should represent the actual
terrain where any structures above the terrain have been removed.
Bicubic should in theory produce sharper edges than bilinear (you
could compare bilinear and bicubic results of r.proj or i.rectify,
although the bilinear/bicubic interpolation methods used by
r.proj/i.rectify differ from the methods used by v.surf.bspline),
maybe that is the reason why the wiki page recommends bicubic for DSM
interpolation.
>
>
> Also, the wiki page and man pages for v.lidar.* cite examples of lidar
> data with two returns. How well should the commands run with 4
> returns? Any big changes
> in th procedure explained in the wiki?
>
It is possible that the lidar toolchain does not work with 4 return
because it classifies points into first or last return, nothing in
between. In theory, it would suffice to identify the first and the
last return, independent of the number of returns, in order to get a
DSM and a DTM. The lidar filtering toolchain should work if you could
remove 2. and 3. returns first.

HTH,

Markus M