RE: [Algorithms] fast triangle-segment test *with precomputation*

[Alex P. <al...@Ex...>]

A few questions about your 3rd test:
- Are you testing line or line segment intersection?
- Why do you need to use the diagonal?
- Doesnt the test hold for any bbox edge as well.
- Also why are you testing 8 lines?  Four parrallel
  bbox edges should be enough to cover any case where
  a corner 'punches through the triangle'

Alex

-----Original Message-----
From: SHA...@ao... [mailto:SHA...@ao...]
Sent: Saturday, August 05, 2000 4:55 AM
To: gda...@li...
Subject: Re: [Algorithms] fast triangle-segment test *with
precomputation*


In a message dated 05/08/00 01:53:24 !!!First Boot!!!,=20
Chr...@Pl... writes:

<< That only works as a conservative test, but not as an exact test.
=20
 If the box, straddling the triangle plane, is just outside one of the
 vertices of the triangle, then both criteria are met, but there's no
 actual intersection. >>

I am currently doing a triangle/aabb test for an octree implimentation.
This=20
test is as near as exact as I can think of. It does 3 tests, each one
more=20
time consuming than the last...

1) Are any of the 3 tri vertices inside bbox...if so return true
2) Does any tri edge intersect bbox face...if so return true
3) Generate 8 lines, each one connects a corner of the aabb to it's
diagonal=20
opposite then test, does this line intersect tri face...if so return
true
return false.

The last test is in case any corner of the aabb punches through the face
of=20
the tri, in which case the tri is technically inside the aabb.

How do you do yours??

John.

_______________________________________________
GDAlgorithms-list mailing list
GDA...@li...
http://lists.sourceforge.net/mailman/listinfo/gdalgorithms-list