[GRASS-dev] New raster-to-raster3d conversion module

Mon Jul 10 18:52:40 EDT 2006

Dear Benjamin and devs,

On Monday 10 July 2006 17:07, Benjamin Ducke wrote:
> Soeren,
> 
> that sounds cool.
> However, I am to dumb to understand just exactly what the module
> does from the screenshot you provided.
> Could you elaborate?

I will try, but my "d"english is horrible.

With r.elev.to.rast3 you can put raster map values from a specific map (input) 
in a specific elevation layer within a g3d volume, based on a 2D elevation map (elev).
Every 3D cell of the volume that intersect with the elevation map will be set to the cell value of the input map.
If the hight value of the elevation map is exactly located at the border between two 3D cells, both cells
will get the input map value.

This works perfectly for putting geological strata into g3d volumes. I need this for modelling
a 3D aquifer.

The screenshot 
http://www-pool.math.tu-berlin.de/~soeren/grass/modules/screenshots/r.elev.to.rast3.png
shows 3 options how to put raster map values into a g3d map.

Here are the options of r.elev.to.rast3:

Description:
 Creates a 3D volume map based on a 2D elevation and value raster map

Usage:
 r.elev.to.rast3 [-ul] input=string elev=string output=string
   [upper=value] [lower=value] [--overwrite]

Flags:
  -u   Use the input map values to fill the upper cells
  -l   Use the input map values to fill the lower cells
 --o   Force overwrite of output files

Parameters:
   input   2d raster maps which represent the values
    elev   2d raster maps which represent the elevation
  output   output 3dcell map
   upper   The value to fill the upper cells, default is null
   lower   The value to fill the lower cells, default is null

In the upper volume of the screenshot the 3D elevation layer based on the grey cutplane 
on the left was filled with the data from the (colored) 2D input map. Also all
the 3D cells above the elevation layer are filled with the values from the input map.

In the middle volume only the elevation layer was filled with the input map values, upper
and lower 3D cells are not filled (null).

In the lower volume the elevation layer and all 3D cells lower than the elevation layer are filled
with the input map values.

> 3D functionality in GRASS is really getting somewhere.
> The only things I could wish for:
> 
> 1. Make volume vizualization work in NVIZ

Well, it works for isosurfaces and cutplanes. But i dont know how to implement real
volume rendering. (have no clue of OpenGL) 
For this reason i implemented the r/r3/v.out.vtk export modules to visualize all
the grass data with VTK based programs like Paraview or MayaVi.
And VTK provides much more visualisation capabilities you can imagine. We will never have 
the manpower to implement this into nviz.

btw.:
VTK provides Python bindings. Thus with grass swig/Python support and the VTK Python bindings,
direct data interaction/visualisation can be implemented. Hm, maybe i should learn Python .... .

> 2. Module to convert 3D Polygon -> 3D raster map

You will need something like a volume rasterizer to do this. That is not easy to implement.
And the hardest thing of 3D polygons with more than 4 vertices is, that the polygon surface is not
well defined. You will need additonal points for triangulation. Otherwise you dont get the surface you expect.
Only triangles or quadrangles are well suited for rasterisation.

But r.elev.to.rast3 provides limited potentialities for this.
You can convert lines, boundaries or areas into raster maps and put them with r.to.rast3 or r.elev.to.rast3,
into a g3d volume.

> 3. Module to query 3d raster value using 3D vector points/coordinates

You can do this with Paraview. VTK will give you the options to combine/intersect 
pointdata, polydata or unstructured grid data with any kind of volume data.
Paraview will give you the data on specific points or a specific path within the volume.
You can use threshold filter, grid filter, cutplanes and much more to discover your 3D data.

> 4. Categorization and information like r.report provides

I was thinking today about a module like r.stats -> r3.stats. But i'm just thinking about it. ;)

> 
> Maybe for GRASS 6.2?

You are welcome to help implementing this. :)

Best regards
Soeren

btw.:
r.elev.to.rast3 is in CVS now. But it will not build by default, the entry in the Makefile
is "missing". 
A new grass test suite with r.elev.to.rast3 tests is available (version 0.2.0.9):
http://www-pool.math.tu-berlin.de/~soeren/grass/GRASS_TestSuite/GRASS_Testsuite-0.2.0.9.tar.bz2

> 
> Thanks,
> 
> Benjamin.
> 
> Soeren Gebbert wrote:
> > Dear Hamish,
> > 
> > On Monday 10 July 2006 02:53, Hamish wrote:
> > 
> >>Soeren Gebbert wrote:
> >>
> >>>i would like to add a new raster-to-raster3d conversion module
> >>>to the grass6.1 CVS in the near future.
> >>>
> >>>The module is called r.elev.to.rast3. 
> >>>It creates a new 3D raster map from a 2D elevation and a 
> >>>2D "value" raster map. The principal functioning is shown here:
> >>>http://www-pool.math.tu-berlin.de/~soeren/grass/modules/screenshots/r.elev.to.rast3.png
> >>
> >>Congratulations on the new module!
> > 
> > 
> > Many thanks!
> >  
> > 
> >>>I would like to ask if i can commit this module next week, or if i
> >>>have to wait until  GRASS 6.1.0 is released? 
> >>>I have tested this module with different datasets and it seems to work
> >>>properly. And i will add some tests to the grass test-suite.
> >>
> >>
> >>If the new module is ready to be in CVS, but you are not sure if it is
> >>release-ready, you can always put it in CVS but not add it to the
> >>Makefile or GUI menus. (as long as the change doesn't touch other parts
> >>of the software)
> > 
> > 
> > Ok, i will submit it after some more testing, but will not add it to the Makefile/GUI.
> >  
> > 
> >>>I think together with r.to.rast3
> >>>http://www-pool.math.tu-berlin.de/~soeren/grass/modules/screenshots/r.to.rast3.png
> >>>r3.to.rast
> >>>http://www-pool.math.tu-berlin.de/~soeren/grass/modules/screenshots/r3.to.rast.png
> >>>and r3.cross.rast
> >>>http://www-pool.math.tu-berlin.de/~soeren/grass/modules/screenshots/r3.cross.rast.png
> >>>the raster<->rast3d modelling/conversion capabilities are mostly
> >>>complete.
> >>
> >>
> >>Small versions of those pics would be a great addition to the help
> >>pages, see r.terraflow, v.voronoi for examples.
> > 
> > 
> > Thats a great idea, i will do this.
> > 
> > Best regards
> > Soeren
> > 
> > 
> >>
> >>Hamish
> >>
> >>
> > 
> > 
> > _______________________________________________
> > grass-dev mailing list
> > grass-dev at grass.itc.it
> > http://grass.itc.it/mailman/listinfo/grass-dev
> > 
> > 
> 