[GRASS-dev] v.what.rast support for centroids, v.rast.stats speed, r.category update

[Michael Barton]

On 10/9/07 5:42 AM, "Hamish" <hamish_nospam at yahoo.com> wrote:

> Hi,
> 
> any comments re. the following patch?
> 
> it lets you use v.what.rast for centroids as well as points.
> 
> I have made a (100m) buffer around a number of points, and in addition to
> the raster stats for the buffer provided by v.rast.stats I want to know
> the exact value at the center point (field sampling station). In this
> case my centroid is exactly in the middle of the area so it's ok. I'm not
> so sure that this patch is useful when the centroid is randomly placed
> within the area, hence this email. Then again the result is actually
> uploaded to the centroid, not the area, so maybe all-correct just
> sometimes requires a grain of salt in the interpretation.
> 

This brings up an issue that we might want to think about for GRASS 7.
Currently, centroids must always be created manually to transform a closed
polyline into an area. While we might want to keep this as an option (but
maybe not), I think it would be better to have some kind of 'area-creation'
utility that would transform a close polyline into an area by automatically
adding a centroid at a predefined center location (geometric center or other
measure) and adding a cat value. Then analytical utilities like you are
discussing here could be more reliable.


> 
> 
> Another issue touched by a recent thread* re. bounding boxes of features
> within a vector map: v.rast.stats is VERY slow. This is because it creates
> a r.mapcalc MASK at the full raster region resolution for each cat it
> processes. When picking out small vector buffers from a large raster this
> is highly inefficient. It would be much better to find the bounding box
> of each selected vector cat, zoom in on it with the GRASS_REGION or
> WIND_OVERRIDE enviro variables for the duration of the r.mapcalc and
> r.univar calls for each cat.
> 
> [*] http://thread.gmane.org/gmane.comp.gis.grass.user/20327
> 
> I would be cool to have a C module which could do 'v.info -g' bounds for
> cats=1,3,5-9,13 &/or where='SQL query'.  ??
> 
> 
> Hamish
> 
> ps- r.category label setting updates are mostly done, it just needs some
> cleaning. I expect it will need more testing and debugging than most
> changes as the labeling lib fns are very dusty.

Just want to say thanks for these new extensions to r.cat -> r.category.

Michael

__________________________________________
Michael Barton, Professor of Anthropology
Director of Graduate Studies
School of Human Evolution & Social Change
Center for Social Dynamics and Complexity
Arizona State University

phone: 480-965-6213
fax: 480-965-7671
www: http://www.public.asu.edu/~cmbarton