tips on using multiple cameras to improve accuracy?

Hi,

I've succesfully used ARToolKitPlus to track multiple devices where each
device has been defined by a multi-marker config.

In the current configuration I am using only from 1 camera viewpoint
from which the images are captured. The pose estimation is used to
render an image from a virtual camera viewpoint somewhere in 3D space
(so not at the same location as the real camera). As discussed earlier
on this list this results in a pose estimation with inaccurate depth
information about the markers/devices.

To improve the accuracy further I am going to use multiple cameras which
look at the markers/devices from multiple directions. Therefore, my
first idea is to define multiple trackers, one for each camera, and then
combine the pose estimations from all the cameras. Straightforward idea
is to use triangulation techniques as the combination algorithm.

But first I would like to read your opinion about this idea, before I am
starting to implement this. Can you help me?

thanks a lot already!
bye
Raymond

-- 

 Raymond de Vries, VR Specialist
  Insight Services
  SARA Computing and Networking Services, PO Box 94613, 1090 GP  AMSTERDAM
  email: raymond at sara.nl   Phone +31-20-5923000  Fax +31-20-6683167

From:	Raymond de Vries <raymond@s ......>	Received:	May 30, 2006
To	ARToolKit Mailinglist <artoolkit@h ..................>
Subject:	tips on using multiple cameras to improve accuracy?
Hi, I've succesfully used ARToolKitPlus to track multiple devices where each device has been defined by a multi-marker config. In the current configuration I am using only from 1 camera viewpoint from which the images are captured. The pose estimation is used to render an image from a virtual camera viewpoint somewhere in 3D space (so not at the same location as the real camera). As discussed earlier on this list this results in a pose estimation with inaccurate depth information about the markers/devices. To improve the accuracy further I am going to use multiple cameras which look at the markers/devices from multiple directions. Therefore, my first idea is to define multiple trackers, one for each camera, and then combine the pose estimations from all the cameras. Straightforward idea is to use triangulation techniques as the combination algorithm. But first I would like to read your opinion about this idea, before I am starting to implement this. Can you help me? thanks a lot already! bye Raymond -- Raymond de Vries, VR Specialist Insight Services SARA Computing and Networking Services, PO Box 94613, 1090 GP AMSTERDAM email: raymond at sara.nl Phone +31-20-5923000 Fax +31-20-6683167