CGAL users discussion list

[Richard Downe <>]

I appreciate the insight.
Yes, when it fails it terminates very fast.  I had seen the 20 point 
initialization, and had lifted the value verbatim from the sample code; 
I will try increasing the number, though I can definitely provide a 
traits class with more consistent initialization points.

One question, though: if I derive my own traits class, should the points 
be on the surface, inside the enclosed volume, or a combination of the 
two, such that the generated tetrahedra intersect the surface to be 
resampled?
--rd

On 08/15/2011 04:32 AM, Michael Hemmer wrote:
> Hi,
>
> I am not the expert for the package but, AFAIK, the experts are on
> vacation. Here is my guess:
>
> In the beginning of the process the point on the surfaces are
> generated randomly by shooting random rays. Usually these are 20
> points [1]. Sometimes this is not enough. It may happen that all
> generated tetrahedra are classified as outside.  In this case, no
> facet is a boundary facet, which in particular means that there are no
> bad facets and the refinement process does not even start. It would
> also mean that your process terminates very fast. Is that correct ?
>
> Increasing the number of these initial points may be already
> sufficient for you [1].
> If this does not help, it may be necessary to chose the initial points
> more carefully. That is, you provide your own traits class, based on
> the one that you are using at the moment, but with your own
> implementation of SurfaceMeshTraits_3::Construct_initial_points [2].
>
> hope that helps,
> Michael Hemmer
>
>
> [1] 
> http://www.cgal.org/Manual/latest/doc_html/cgal_manual/Surface_mesher_ref/Function_make_surface_mesh.html#Cross_link_anchor_1460
> [2] 
> http://www.cgal.org/Manual/latest/doc_html/cgal_manual/Surface_mesher_ref/Concept_SurfaceMeshTraits_3.html#Cross_link_anchor_1472
>
>
>
>
> On Mon, Aug 15, 2011 at 5:27 AM, Richard 
> Downe<>
>   wrote:
> > In running surface mesh resampling using the example taken from the
> > Polyhedron demos in CGAL 3.8, I normally get a good result, but with
> > alarming frequency and apparently nondeterministic cause, I sometimes get an
> > empty set (e.g., the C2t3 contains 0 facets after reconstruction completes).
> >   Sometimes running the program a second time causes a correct result to
> > emerge.  Is there anything I can do to tune the algorithm to ensure that I
> > always get a valid result?
> > --rd
> >
> > --
> > You are currently subscribed to cgal-discuss.
> > To unsubscribe or access the archives, go to
> > https://lists-sop.inria.fr/wws/info/cgal-discuss