[vtkusers] ThinPlateSpline regression

[David Gobbi]

Hi Michael,

No, I don't think that ThinPlateSplineTransform provides a way to do this.
I see a ComputeBestFittingPlane() method in vtkDelaunay2D, but it is a
protected method and, looking at the method that it uses, I don't think it
computes the "best" plane, anyway.

Hopefully other people who have used VTK to compute the best-fit plane
can give you a better answer, but my suggestion is to search the internet
for a method that computes the best plane via an eigenvalue decomposition
(my suspicion is that for the best-fit plane, the eigenvector for the smallest
eigenvalue is the normal of the plane, and the "origin" of the plane is the
centre-of-mass).

The vtkMath::Jacobi() function can compute the eigenvalues.

 - David


On Fri, Aug 29, 2014 at 3:29 AM, Brehler, Michael
<m.brehler at dkfz-heidelberg.de> wrote:
> Hi,
>
>
>
> I am using VTK to find a best fitting hyperplane (or regression plane) for a
> given 3D point cloud (polydata surface) and was wondering if the
> ThinPlateSplineTransform could be of any help. Can I use an equally
> distributed 'grid' of points to calculate the matching hyperplane with the
> ThinPlateSplineTransform or is there a better way?
>
>
>
> Regards,
>
> Michael