I have to align 4 3d coordinate systems: real world, projected geometry (pseudo-holography with 3 beamers), 2 kinects (skeleton tracking). Till now I’ve been doing this manually and as you can imagine, it’s a pain in the #ss.

While aligning the projections to the real world would remain a manual process I would like to automate the alignment of the 2 kinects’ spaces to the projected geometry. I think this should be possible by calculating the transform matrices between the same 3 points represented in the projected geometry’s space and each of the kinect’s spaces, then align the kinects’ tracked data by using that transform matrix.

Does anyone know the math to do such a thing? Or resources where to find it? I didn’t find it googling. Perhaps cause I’m using wrong terminology.

]]>https://groups.google.com/forum/#!topic/simple-openni-discuss/e6Ag6IwMEuQ

Though I haven’t achieved very precise results with it yet. Trying to work out a best practice now.

]]>