You should keep in mind that triangulation might work very differently for small data sets (i.e. simple polygons) and large data sets such as Kinect point clouds.
You might want to look into QHull
http://www.qhull.org/, it's complicated but has features for convex hull and Delaunay triangulation. I seem to recall that someone did an experiment with integrating QHull and Processing, but can't find the link right now.