After finishing the triangle geometry conversion, I have restarted looking at the Triangulate3D function. For some crazy reason, I still can't figure out why IsConvex() doesn't work quite right. Or rather, I know why it doesn't work right, but I am unsure how to fix it. The problem lies in the situation where when I am inserting consecutive points into a csPlane3 object, sometimes the resulting plane is defined counter clockwise, and sometimes it is defined clockwise. e.g. if one considers the polygon defined by the points:
(0, 10, 10)
(0, -10, 10)
(0, -10, -10)
(20, 5, -10)
(10, 0, -10)
(0, 10, -10)
Point number 4 (that is (10, 0, -10)) should be convex. However, what happens is that since the polygon loops back around, the plane defined by points 4, 5, and 3 is defined counter-clockwise, resulting in an incorrect result for the IsConvex() test. I am not quite sure how to fix this. Anyone who thinks they can provide some insight, feel free to look at my code. It's available in the libcrystalspace project (in MSVC) under the files triangulate3d.h and triangulate3d.cpp. It's been placed in the csgeom file, if you are not using msvc.
As a result of this problem, I am going to continue with the collada conversion by next moving to materials conversion. The fact that the collada conversion system doesn't currently support any polygons other than triangles isn't immediately a problem, since COLLADA exporting from Blender allows the user to export only triangles, which can then be used to test the collada system. I will look into the materials conversion today, and hopefully have it coded in the next few days, at which time I will re-investigate the triangulation function.
No Trackbacks/Pingbacks for this post yet...
Comments are not allowed from anonymous visitors.
|<< <||> >>|