Project: ARToolKit with user invisible markers

[djoele]

We have a project running here that tries to develop an AR system with user invisble markers. These markers will be written in ink that is only visible in the infrared range. We will use an infrared camera to detect the markers. We are also going to use a video see-through HMD. We have not gon into details yet.
We wil use ARToolKit 2.70.1

What I think is the following:

First the camera and markers are initialized.

There will be a mainloop:
The infrared camera is used to detect the marker. Then the transformation matrix is calculated. This transformation matrix is used to draw virtual objects in the micro displays of the HMD. (Of course some manipulation needs to be done on the matrix)

Now I have some questions:

Do you think this approach will work? Or do you have any additions or other ideas?

Does ARToolKit 2.70.1 provide functions to displays images and virtual objects into the microdisplays of an HMD? Which functions are that?

How is the HMD treated by ARToolKit? Is it considered as 2 cameras connected to a computer or is it considered as just an HMD? So, will ARToolKit consider 3 cameras to be connected to the computer?

How will everything need to be connected? THe infrared camera is connected
to a USB port or Firewire? How to connect the HMD? Do they have two wires coming from the cameras from the HMD, so they can also be connected to USB ports? Or is there another way?

[djoele]

A video see-through HMD is treated by ARToolKit as just 2 cameras connected to a computer? Or how to connect an HMD? Anyone has done that
before and can give me some advice?

[djoele]

Now I have a question on the translation matrix from ARToolKit.

We want to use one camera to detect the markers and from that calculate
the transformation matrix. Then we want to insert virtual objects in the image
of another camera. But now the problem is to convert the obtained matrix into a correct matrix for the second camera.

I read that ARToolKit uses a 3*4 matrix for the transformation matrix, T.
The marker is fixed, a 3*1 vector M=(Xm,Ym,Zm) Then the camera coordinates, a 3*1 vector C=(Xc,Xy,Xz) are calculated by ARToolKit as:

C = T*M

Now we have 2 cameras, one located at (x,y,z) and the other one at (x+delta,y,z). We want to use the first camera to calculate the transformation matrix for the second camera.

How to do this? Is it simply using the matrix of the first camera and changing the values of the fourth column?So, use (x+delta,y,z) instead of (x,y,z).