CGAL users discussion list

["Laurent Rineau (CGAL/GeometryFactory)" <>]

Le lundi 02 juillet 2012 17:37:13 Hugo Mercier a écrit :
> Le 29/06/2012 16:12, Sebastien Loriot (GeometryFactory) a écrit :
> >> If I want to test the intersection with a >=2-dimension primitive, I
> >> think a good way is to build a Polyhedron_3 from that primitive and
> >> substract it from the Nef_polyhedron.
> >> 
> >> But what about points and lines ?
> > 
> > You should use this [1] package for that purpose.
> > The idea is to reduce intersection test at the face level.
> > 
> > (Note that you can use a polyhedron instead of a nef-polyhedron).
> > 
> > You can also test the intersection between two polyhedra with this
> > package.
> 
> Thank you for your answer.
> 
> How is it related to AABB trees ?
> I am currently using them to speed up intersection tests. I understand
> that "Box_intersection_d" are more general than AABB trees (especially
> in the bipartite case), but how do they behave, for instance, on the
> case of a 1-triangle-N-segments intersection test where do_intersect(
> triangle ) is called on an AABB tree built from a set of Segment_3 ?

"AABB Tree" and the "Bbox_intersection_d" are different packages, with 
different approaches.

"AABB Tree" is for intersections tests when you have a big object that is 
fixed, and a set of intersections queries between the big object and small 
query objects. the AABB Tree stores a hierarchy of bounding boxes for the big 
object.

The "Bbox intersection" package is for one-shot intersections queries for
  - self-intersection of a big object,
  - or for intersections of two big objects.
No data structure is kept in memory between two different calls to functions 
of that package.

-- 
Laurent Rineau, PhD
R&D Engineer at GeometryFactory           http://www.geometryfactory.com/
Release Manager of the CGAL Project       http://www.cgal.org/