[IGSTK-Developers] Some component design questions

[Andinet Enquobahrie]

Hi Kevin,

Let me try to answer some of the questions.

>
> 2. The Landmark3DRegistration state machine forces 3 or more points to
> be selected for the registration process. The state machine does not put
> an upper bound on the number of points used. Is there a theoretical
> upper bound? Is there an upper bound commonly used in practice? Does
> including more points reduce the error significantly?


The landmark3DRegistration class internally uses the ITK landmark
registration
initializer which is basically implementation of the closed-form solution to
the least
squares problem developed by Horn et al. For your reference, here is a link
to their
original paper.

http://people.csail.mit.edu/bkph/papers/Absolute_Orientation.pdf

>From the algorithm description,  there is no upper bound to the number of
landmark
points. However, they didn't provide any error analysis that shows an
increase
in the number of points reduce the error "significantly".


3. The tracker's "AttachObjectToTrackerTool" method is public (not
> called by anyone) and has no interaction with the state machine. It is
> typically called by applications and results in sending a request to the
> SpatialObject::RequestAttachToTrackerTool method. Should there be a SM
> check here? Will the Tracker allow such an assignment in all states?


This method should have been routed through the SM. In any case, this method
will be deprecated in the new Tracker interface


-Andinet
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/igstk-developers/attachments/20071024/d986aac7/attachment-0001.html>