In this video I'm demo-ing the kinect and testing a concept for tracking.
This uses a core image filter to threshold out objects depending on nearness to the kinect. Then, a clever version of a centroid calculation routine by smokris and jstrecker of kineme (kineme.net) that was adapted from GPU Gems (http.developer.nvidia.com/GPUGems3/gpugems3_ch26.html ) is used to place the sphere object in x/y translation.
I've added to this routine to make the object of interest disappear from screen if nothing is currently crossing a specified z threshold.
So this is a way of doing many of the same activities that people use OpenCV, haar detectors, blob tracking, tuio, ofx, processing, etc., for.