From: Christoph F. <c.f...@cg...> - 2005-08-26 10:09:31
|
Marcus Lindblom wrote: >>> So I'd say I expect them to be functional, but I can't promise that >>> they're bug-free. >> > > They're not even functional. :-/ > > Actually, they're horribly buggy and quite unusable. (Have tested > sphere<->cyl and box<->cyl. Converting my cylinders to AABB and > testing produces valid and reasonable results (although not very > useful in the application). So, I'm quite confident in my geo->cyl and > cyl->geo functions, but for cylindervolume-intersection tests, I get > false positives and false negatives all over the place. > > Using OOBB's would be acceptable here for me, but they are not available. > > I'll dig up some cyl<->sphere algorithm from the net and see how that > works out, otherwise I'll be forced to abandon the cylinder approach > altogether. > > Best Regards > /Marcus What does the Cylinder Sphere intersection test do in OpenSG, did you look into it? Is it conceptually wrong, or does it have bugs? For the cylinder sphere part there is an article from JGT about Point Cylinder Distance (with code) see http://www.acm.org/jgt/papers/BarbierGalin04/ I also don't think OBBs are an option, as they are difficult to update under (large) deformations. Regards, Christoph |