[vtkusers] Point cloud downsampling/voxelization

[Christian Jauvin]

Hi,

I am currently building a LIDAR-acquired, botanical tree point cloud
analysis tool, which includes, among other aspects, reconstruction
algorithms.

The visualization module of the current version is based on a set of
simple primitives: the point cloud itself (as a polydata), and a set
of sphere actors, with which the user interacts.

I'd like to offer a mode where the point cloud would be downsampled to
a set of cubic bins, or voxels. What I have in mind is very similar to
this PCL mechanism:

http://docs.pointclouds.org/trunk/classpcl_1_1_fast_voxel_grid.html

Ideally, I would like this mode to retain the same interactivity I
have with my current mechanism, with which I can, for instance, use a
vtkSelectEnclosedPoints to find the points lying inside a sphere, and
act on them accordingly (i.e. retrieving their id, or change their
color).

I'm aware that simply using a large number of cube actors would be
prohibitive, so what I have in mind would be similar to my current
point cloud representation (i.e. a single polydata), but cube-based
instead of point-based.

I've been scouring the mailing lists and tutorials to find an example
that would remotely resemble what I'd like to do, to no avail, so any
help or hint would be greatly appreciated.

Christian