Hello,
I have a set up in which 6 persons are wearing videogoggles on which a
firewire-camera has been fixed.
I need to track the position of the head/camera of each person; the movement
of the person is limited to the movement of the head (as if sitting on a
chair).
We now use magnetical tracking (for a one person setup), which gets too
easily biased (calibration) and is too complex for our particular
application. 'Classical' optical tracking systems are too expensive.
Could this be done with ARTools?
If the camera would be seeing marker cards in various XYZ-positions in an
environment of 1 to 3 meters wide, would it then be possible to calculate
the relative positon of the camera itself, rather accurately?
I understand ARTools has not been designed for this use, but it looks like a
logical extension.
On the list it appears comparable, but more complex experiments have been
carried out (mobile environments,..)
If ARTools need to be (realistically) altered, I have the assistance of a
medialab.
Thanks for an answer!

Eric

eric.joris@s ........
eric.joris@c .............
http://www.kaufhausinferno.com
http://www.man-o-war.be
http://www.crewonline.org

Hi,
    I have tried what you are suggesting, it produced rather poor results,
however I was only using one marker. It is possible to improve the results
using multiple markers and biasing them depending on certain factors such
the amount of pixels they project onto. I found that there seemed to be
considerable jitter, this can also be minimised with the use of a kalman
filer or extended kalman filter. My advice would be to develop a hybrid
system using inertial sensors as well, while using artoolkit to correct for
drift. Well that's my take on the problem.
Best of Luck

Jean-Daniel Nahmias UCL

----- Original Message ----- 
From: "eric skyline" <eric.joris@s ........>
To: "artoolkit" <artoolkit@h ..................>
Sent: Wednesday, January 07, 2004 10:32 PM
Subject: Using ARTools for optical tracking


> Hello,
> I have a set up in which 6 persons are wearing videogoggles on which a
> firewire-camera has been fixed.
> I need to track the position of the head/camera of each person; the
movement
> of the person is limited to the movement of the head (as if sitting on a
> chair).
> We now use magnetical tracking (for a one person setup), which gets too
> easily biased (calibration) and is too complex for our particular
> application. 'Classical' optical tracking systems are too expensive.
> Could this be done with ARTools?
> If the camera would be seeing marker cards in various XYZ-positions in an
> environment of 1 to 3 meters wide, would it then be possible to calculate
> the relative positon of the camera itself, rather accurately?
> I understand ARTools has not been designed for this use, but it looks like
a
> logical extension.
> On the list it appears comparable, but more complex experiments have been
> carried out (mobile environments,..)
> If ARTools need to be (realistically) altered, I have the assistance of a
> medialab.
> Thanks for an answer!
>
> Eric
>
> eric.joris@s ........
> eric.joris@c .............
> http://www.kaufhausinferno.com
> http://www.man-o-war.be
> http://www.crewonline.org
>
>
>

Thanks Jean-Daniel for your answer.
How than about this possibility:
Suppose we simply fix a camera on the ceiling, looking down upon the
viewers' head (close up). The viewer wears a marker card upon the head
(eventual a second marker and a second camera on the side of his face)
Again, the person only moves his head.
Wouldn't we able to get a rather accurate tracking this way?
Thanks again,
Eric
> From: "Danny Nahmias UCL" <d.nahmias@c ...........>
> Date: Thu, 8 Jan 2004 00:10:32 -0000
> To: "eric skyline" <eric.joris@s ........>, "artoolkit"
> <artoolkit@h ..................>
> Subject: Re: Using ARTools for optical tracking
> 
> Hi,
> I have tried what you are suggesting, it produced rather poor results,
> however I was only using one marker. It is possible to improve the results
> using multiple markers and biasing them depending on certain factors such
> the amount of pixels they project onto. I found that there seemed to be
> considerable jitter, this can also be minimised with the use of a kalman
> filer or extended kalman filter. My advice would be to develop a hybrid
> system using inertial sensors as well, while using artoolkit to correct for
> drift. Well that's my take on the problem.
> Best of Luck
> 
> Jean-Daniel Nahmias UCL
> 
> ----- Original Message -----
> From: "eric skyline" <eric.joris@s ........>
>> Hello,
>> I have a set up in which 6 persons are wearing videogoggles on which a
>> camera has been fixed.
>> I need to track the position of the head/camera of each person; the
> movement of the person is limited to the movement of the head.
>> Could this be done with ARTools?
>> If the camera would be seeing marker cards in various XYZ-positions in an
>> environment of 1 to 3 meters wide, would it then be possible to calculate
>> the relative positon of the camera itself, rather accurately?
>> Eric