[vtkusers] Re: VTK and transformation

[Andrew Dolgert]

Would you believe there are three separate interfaces to vtkProp3D, the
base class for vtkActor, which all allow you to transform an actor?  In
order to assign actor transformations based on haptic input, you poke
the transformation into vtkActor, or, rather, its base class vtkProp3D.
Then you can use a vtkCellPicker to find what point is picked.
vtkCellPicker knows how to apply these transformations in reverse.
1. You can take your tranformation from haptic input and set the
vtkActor's position and orientation directly, using converstions you
must already have seen in vtkTransform or vtkMatrix4x4.  You would do
this at every timestep.
2. You can take your transformation, in whatever form you have it, and
set the actor's UserMatrix property directly.  This can be a shorthand
way of doing the same as setting position and orientation separately,
although the two choices compound each other.
3. If you call vtkActor::SetUserTransform, then this works slightly
differently.  One difference between a vtkTransform and a vtkMatrix4x4
is that a vtkTransform remembers where it came from.  It is a live link.
If you call SetUserTransform with a pointer to the same vtkTransform as
that set by your haptic input, or even a pointer to a vtkTransform
derived from your haptic input by chaining transforms, then the vtkActor
will be updated for you every time you get more haptic data.
For an example of how to convert from the vtkActor's coordinates to
world coordinates, you might look in vtkPicker::Pick()'s implementation,
or you might let vtkCellPicker do the work for you if that's possible.
(It would be with some modifications on your part to excise the 2D
mousing code.)  The best advice I've seen on how to do the conversions,
aside from reading Feiner, van Dam, et al, is from the VR Juggler
guide's intro to matrices, which suggests writing matrices in the form
w_M_a to represent a matrix translating from actor space to world space,
in order to keep your transformations straight.  Instead of using the
terminology of model, view and perspective transformation matrices, VTK
has actor coordinates, world coordinates and screen coordinates.  It's
like this:
 screen coords - s_M_w - world coordinates - w_M_a - actor coordinates
where the user transform you put in the actor is a transformation from
the world to the actor frame, a_M_w, and the matrix s_M_w is the
vtkCamera's "composite perspective transform."  This is a combination of
what graphics people call the view and perspective transformation
matrices, and vtkCamera also offers you what is usually called the view
matrix in its "perspective transform."
Drew Dolgert
Cornell Theory Center
 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.vtk.org/pipermail/vtkusers/attachments/20031028/f9c1fe8f/attachment.htm>