Point cloud to STL file

From: Anshuman Razdan (razdan@taurus.eas.asu.edu)
Date: Tue Jan 20 1998 - 07:14:07 EET

Unless I dont understand the definition of point cloud data, I dont
understand why one would go from an stl file to pointcloud data
and then to a solid model.

It is true that stl file has only geometry and no topology (i.e. connectivity
information between different entities and things like holes, slots etc.).
But there is a chance to convert the faceted (STL) model to a topologically coherent
solid model. However, once you convert to point cloud data (either 3 points
of the triangle or one point per triangle - which is even worse) you
lose total ability (theoretically) to recover the topology.

Point Cloud data to surface data: There has been considerable work
over the past few years in the CAGD (Computer Aided Geoemtric Design) arena
to convert point cloud data (for example from a digitizer) to surface
such as B Spline and then to solid model. Many different methods have been
applied such as Alpha Shapes etc. (More info can be found at the follow
web pages:)

Not to inundate you all with CAGD stuff but suffice to say that
when trying to put surfaces to pure point cloud data there is ambiguities
as to where hole boundaries start and end.

OK - so enough on the stl->pt cloud->surface->solid model.

I am not sure why one cannot go from stl file to solid model route - via
the surface route. I do have some sowftware but the results are not
publishable as yet. If any of you do want to support academic research/soln
in that area, you are welcome to contact me (shameless pandering but thats
academician's basic right). Afer all surface fitting to triangles is much
better than fitting to point clouds.

It is also worth noting that the kind of surface you expect can have a major
impact on the surface/solid solution if you have free form surfaces. The
additional factor of C/G continuity - the type of continuity for the
surface. C1 (tangent continuous) vs C2 (curvature continuous)
solution for the surface can result in different solutions. But this is
only for the case where free form surfaces are involved (which the modern
design world is fast moving to).

Most commercial software (this is not a critique so
no flames pl) makes you "pick" curves out of the point cloud data and
it will then skin a surface through these curves given some tolerances etc.
Software such as Imageware and raindrop's geomagic etc. deal with this
kind of stuff - I am not sure how holes might be automatically detected in these
software though (Does anybody know if they do - I have my doubts).
Dr. Anshuman Razdan
Technical Director PRISM
* Ph: (602) 965 5368 FAX: (602) 965 2910
* Office: GWC 574 Email: razdan@asu.edu
* Snail Mail: PRISM, MCode 5106 Az St. Univ, Tempe AZ 852815106
* http://surdas.eas.asu.edu/~razdan

This archive was generated by hypermail 2.1.2 : Tue Jun 05 2001 - 22:44:44 EEST