Tim Bailey July 12, 2013 GSOC week 4 check in
Horizon based voxel interpolation This was an interesting week. After thorough reconsideration of the foundational documents for this project as well as reviewing about a dozen other three d interpolation projects, I have settled on a methodology that I believe resolves many of the outstanding issues. The vertical to horizontal anisotropy problems are a rabbit hole that I do not need to resolve during this process. Rather than get into the weeds of benefits and liabilities of different interpolation operators, I think that the best way to move forward is by deterministic flood population of voxels. Clearly different disciplines have different methods and needs for spatial approximations. In my estimation, the end users decisions about what the appropriate geostatistical or logical operator to apply should take place between generation of conforming horizons, and the operation that generates the r3 map. The most important point of this is the population of the r3 voxels will proceed through a deterministic operator, which can be accomplished with existing tools. The various approximations such as geostatistical operations can be accomplished before the r3 map is initiated. The r3 map can essentially be flood filled from points by selecting all the points in a tier and applying v.voronoi, followed by v.to.rast. A data artifact from this process is an arbitrary step pattern caused by sparse data. If you want to smooth transitions you can introduce intermediate horizon approximations based on the logical model of the operators choosing. I have a proof of concept dataset which I manually shepherded through what is at this point a ten step transformation between a conforming horizon descriptions, and an r3 map. Next week I plan on continuing working on the generic wrapper that manages the transformation of horizon intervals to voxels. Also I will get my project wiki page up on very soon. Have a good weekend Tim
_______________________________________________ grass-dev mailing list grass-dev@lists.osgeo.org http://lists.osgeo.org/mailman/listinfo/grass-dev