[vtk-developers] Bug or feature in vtkPlane::IntersectWithLine ?

[Will Schroeder]

Philippe-

I agree, if parallel, then a distance check should be performed between the
line and the triangle and compared to the input tolerance. Obviously it's
distance to a *finite line*, so you might start by computing distance of a
infinite line to the infinite plane containing the triangle (this should be
fast). Then if this is within tolerance, then take into account the finite
nature of the line and triangle to return the final verdict.

Similar code should be extended to the polygon as well.

I suspect that this was written as a performance "feature" for picking. The
nature of meshes is such that if a ray is parallel to a triangle t1, then
it is likely to hit another triangle connected to t1 but with a different
normal. And if t1 and all connected triangles are all parallel to the ray,
then you can't see them hence pick them :-) But line intersection is being
used for heavier-duty operations in VTK now, so we should beef the
operation up.

W

On Sun, Nov 27, 2011 at 11:28 AM, Philippe Pebay <philippe.pebay at kitware.com
> wrote:

> Hello all,
>
> The IntersectWithLine()  routine in vtkPlane returns 0, i.e., no
> intersection, whenever the plane and line are parallel.
> Specifically, this occurs in lines 171-20, copied below:
>
>   num = vtkMath::Dot(n,p0) - ( n[0]*p1[0] + n[1]*p1[1] + n[2]*p1[2] ) ;
>   den = n[0]*p21[0] + n[1]*p21[1] + n[2]*p21[2];
>   //
>   // If denominator with respect to numerator is "zero", then the line and
>   // plane are considered parallel.
>   //
>
>   // trying to avoid an expensive call to fabs()
>   if (den < 0.0)
>     {
>     fabsden = -den;
>     }
>   else
>     {
>     fabsden = den;
>     }
>   if (num < 0.0)
>     {
>     fabstolerance = -num*VTK_PLANE_TOL;
>     }
>   else
>     {
>     fabstolerance = num*VTK_PLANE_TOL;
>     }
>   if ( fabsden <= fabstolerance )
>     {
>     t = VTK_DOUBLE_MAX;
>     return 0;
>     }
>
> Nevertheless, a line may be parallel to a plane and yet be contained
> inside said plane. In this case, the routine will return 0 just as if the
> line was parallel and outside of the plane, as the calculation does not
> consider the plane-to-line distance.
>
> It might have been intended to be like this. However, this becomes a
> problem in the implementation of vtkTriangle's implementation of
> IntersectWithLine(), which first calls vtkPlane::IntersectWithLine(): as a
> result, a line which goes through a vtkTriangle, while being contained in
> the plane of this triangle, will be considered as NOT intersecting the
> triangle, which is clearly an error.
>
> I am therefore wondering whether this behavior of
> vtkPlane::IntersectWithLine is a feature, or a bug.
>
> If it is not a bug, then I would submit that the implementation of
> vtkTriangle::IntersectWithLine() should be modified, by checking, for lines
> that ARE parallel to the triangle, whether they are contained in the same
> plane.
>
> Please let me know what you think. Thank you,
> Philippe
>
>
>
>
> --
> Philippe Pébay
> Directeur de la Visualisation et du Calcul Haute Performance
> Kitware SAS
> 20 rue de la Villette
> 69328 Lyon cedex 03, France
> +33 (0)4.26.68.50.03 begin_of_the_skype_highlighting            +33
> (0)4.26.68.50.03      end_of_the_skype_highlighting
> http://www.kitware.fr
>
> _______________________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
>
> Follow this link to subscribe/unsubscribe:
> http://www.vtk.org/mailman/listinfo/vtk-developers
>
>
>


-- 
William J. Schroeder, PhD
Kitware, Inc.
28 Corporate Drive
Clifton Park, NY 12065
will.schroeder at kitware.com
http://www.kitware.com
(518) 881-4902
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/vtk-developers/attachments/20111128/52a9bcdf/attachment.html>