Author Topic: Improving to 6 DOF head tracking  (Read 20870 times)

Offline Timo

  • *
  • Posts: 3
  • Karma: +0/-0
    • View Profile
Reply #15 on: February 06, 2008, 12:56:45 PM
Hi all,

I developed my own tracking system using OpenCV. DeMenthon's POSIT algorithm for 6DOF tracking is part of OpenCV. I tried and tweaked a lot arround with 6DOF tracking and still like POSIT best. Approaches that try to solve mathematic equations directly usually don't work well because of numerical stability.
Unfortunately for my tracking I need 5 points!!! I use a firewire DCAM industrial camera (you can get from 150,-$) and infrared light attached to the camera. On my devices I use reflective markers (usually balls). I can track up to 4 different devices without problem in full 6DOF with good quality. I also use the Kalman filter, that is implemented in OpenCV to reduce the jittering.

DeMenthon's algortithm also works for just 4 points, so it should be possible to use it with the wiimote.
The main difficulty is to relate the 2d points to the 3d-object points. For headtracking, where you can assume the head in a specific position this should be easy.
But DeMenthon also presented a paper SoftPOSIT, were the points do not need to be related. A C implementation of his algorithm is available on his homepage.

The question for me remaining is, how to get the intrinsic camera parameters and if the resolution of the of the wiimote is good enough to give resonable results.

So, does anybody know the wiimote cameras focal point, image point size, resolution and distortion parameters?