[GRASS-user] Mask creation from sporadic data v.hull

[John Tate]

Hi Dylan,

Cheers for this.

I have been playing around with v.category and v.extract this afternoon to 
create a boundary but not getting very far. I still have a lot to learn when 
it comes to vectors and tables in GRASS.

I compared your method r.grow (had to be 4m to get rid of the holes in the 
data) with r.in.xyz at 5m resolution (has to 5m to get rid of all holes) on a 
sample area.

The r.in.xyz at 5m not only had a cleaner edge (no wavy edges around the ends 
of the lidar lines) but seemed to be more accurate in general (less distance 
away from points due to the location of the 5m cell grid).

There will still be some large holes in a few areas doing it this way. Guess 
this will have to suffice for now. I want to try and keep the process simple so 
that it can be automated without any visual interpretation to fill the large 
holes with g.region, g.grow and then r.patch. Or digitising around the whole 
swath.

Thanks for the response though, hadn't used r.grow before so good to learn 
something new.

Cheers

John


On Friday 18 December 2009 16:09:09 Dylan Beaudette wrote:
> Hi,
> 
> I have solved this problem in the past with a combination of g.region,
> v.to.rast, and r.grow.
> 
> For example, if you import your points in as a vector-- convert to
> raster, increase the cell size, and grow out a couple of cells. If you
> are bringing in your lidar data with r.in.xyz, then it is already a
> raster, and you can adjust the resolution and "grow" accordingly. Make
> sense?
> 
> Cheers,
> Dylan
> 
> On Fri, Dec 18, 2009 at 4:04 AM, John Tate <john.tate at ntlworld.com> wrote:
> > Hi all,
> >
> > For a while now I have been trying to make an accurate mask for LiDAR
> > points. As I am working with 1m resolution data I want the mask to be
> > accurate (no more than 1m around the edge of the points/flight area).
> >
> > The data is wavy around the edges, due to a windy flight when the data
> > was collected, so v.hull ignores the waves and joins the crests, adding
> > vast NULL areas to the mask. Also, there are up to 10m holes within the
> > individual swaths.
> >
> > r.in.xyz at 5m resolution has given me the best results so far as final
> > accuracy masks, but not for the processing as I want holes within the
> > data filled with v.surf.rst.
> >
> > I have looked at v.delaunay and v.voronoi too, but both need topology
> > built and don't have similar outputs to v.hull regarding vast NULL areas.
> > This is ok for 1km tiles, but I want to process each swath separately
> > (study on swath accuracy) and these can be c.5million points each.
> >
> > Does anyone know if there is way to do this accurately without manually
> > digitizing? Have I missed something in the book/manual?
> >
> > Cheers
> >
> > John
> >
> > _______________________________________________
> > grass-user mailing list
> > grass-user at lists.osgeo.org
> > http://lists.osgeo.org/mailman/listinfo/grass-user
>