Hey i'm near Gre too.
General recommendation for an algorythm that is very useful for houses and any objects that have flat parts.
Super efficient easy algorythm...
sample grids or spheres comprising 10-50 vertices, measure if they are on the same plane, if they are almost exactly on the same plane, delete 1/2 of them or 2/3rds of them... very good losless mesh simplification algorythm for objects that have straight zones.
nice one thanks.
planar simplification of grid subsample voxels?
-
cosmocompare
- Posts: 10
- Joined: Sat Jan 14, 2017 6:30 pm
-
cosmocompare
- Posts: 10
- Joined: Sat Jan 14, 2017 6:30 pm
Re: planar simplification of grid subsample voxels?
this is interesting: Boissonnat and Cazals [8] introduce a coarse-to-fine
point cloud simplification approach. Their algorithm takes
a random initial subset of the input point cloud and uses
its 3D Delaunay triangulation to define a signed distance
function over the set. This implicit function is then used to
enlarge the initial set until a significant number of points
is found to lie within a user-defined approximation error
tolerance. In the second and final step, the enlarged subset
is Delaunay-triangulated [12] and a surface mesh is reconstructed.
If this initial surface does not meet the error
condition, additional points are inserted iteratively sorted
by their distance to the closest surface facet. The method
thus delivers both a reconstructed and simplified mesh simultaneously.
The algorithm’s point cloud simplification
step is costly due to the construction and maintenance of
a 3D Delaunay triangulation. Given the increasing usefulness
of point-based processing, the algorithm’s restriction
to triangular meshes as output seems undesirable
point cloud simplification approach. Their algorithm takes
a random initial subset of the input point cloud and uses
its 3D Delaunay triangulation to define a signed distance
function over the set. This implicit function is then used to
enlarge the initial set until a significant number of points
is found to lie within a user-defined approximation error
tolerance. In the second and final step, the enlarged subset
is Delaunay-triangulated [12] and a surface mesh is reconstructed.
If this initial surface does not meet the error
condition, additional points are inserted iteratively sorted
by their distance to the closest surface facet. The method
thus delivers both a reconstructed and simplified mesh simultaneously.
The algorithm’s point cloud simplification
step is costly due to the construction and maintenance of
a 3D Delaunay triangulation. Given the increasing usefulness
of point-based processing, the algorithm’s restriction
to triangular meshes as output seems undesirable