How does the Landmark registration get the "from" position

In order to make a realtime EM tracker on the physical model, and the needle followed in the virtual model, I tried the Landmark registration tutorial in IGT.
Firstly, using the vtkLandmarkTransform, I can get the expected transform positions.
But when using the Fiducial Registration Wizard, I don’t know how does the module get the “from” markup points position from a video ? The “to” positions are easy from slicer.

The EM tracker is connected using Plus, the realtime transforms can be seen in the transform module.
so, maybe this is near to the truth. :pray: