On Thu, 2010-11-18 at 06:26 -0800, Prasad Dixit wrote: 
> So, as per my understanding and from your reply... I can mount a single
> camera on the robot (with camera_info topic published) and place various
> patterns at different locations where i know (x,y,z) of all those earlier. I
> don't need ekf or probabilistic approach since i am taking single input from
> webcam (Logitech pro 9000 - upto 30 fps) and know all patterns placing
> positions. A particular pattern will get matched whenever robot comes across
> it and ar_pose will recognize Robot's current position and update it in
> stack.
> Please let me know if i am wrong in this approach.

You may have problems with the Logitech Pro 9000, the C600 may be a
better choice since it has fixed focus and you can swap the lens for a
wider field of view. With both of these however, if you are putting this
on a flying robot you may  have issues a 'jello' effect caused by the
rolling shutter.

We have been using the Point Gray Chameleon cameras because of this.