Diapositive 19 sur 19
Commentaires:
As a conclusion, the first thing I have to say is that we are happy. For our application, the collision detection process is no longer a bottleneck.
More over, the presented method is very easy to implement in an application where OpenGL is used for the visualisation. It needs only a few dozens lines of code which are given in the proceedings.
As OpenGL exists on almost all platforms, there is no portability problem. And as the graphics hardware manufacturers generally provide an optimised implementation of OpenGL for their hardware, we can with no extra effort take advantage of them.
As it requires no pre-computation, this method is particularly efficient for deformable objects.
In our application, we tests for collisions with a triangle mesh, but exactly the same code can detect collisions with more complex geometric primitives, like NURBS for example, as soon as the graphics library implements them.
We can also extend the method to a little more complex objects than our tools. OpenGL provides at least 6 extra clipping planes. So we can detect collisions between a solid made by the intersection of 12 half spaces and any arbitrarily complex object.
Thanks for your attention.
