[vtkusers] How to bring an object to the coordinate's origin

David Doria daviddoria at gmail.com
Mon Mar 28 09:23:41 EDT 2011


On Mon, Mar 28, 2011 at 8:09 AM, nuno.jf <nunofernandes7 at gmail.com> wrote:
> http://vtk.1045678.n5.nabble.com/file/n4267486/1st_good_result.vtk
> 1st_good_result.vtk
> Hi David,
>
> I attached here my first good result, using your implementation of ICP. Once
> again, thank you very much!
> As I will need to write about this version of ICP in my work for college,
> could you please explain me the SetTooFarThreshold function? I know that, by
> using this function, the points that are too far away will not be considered
> to the alignment of the point clouds. My questions are:
> #1. The points that are too far away from what? (Is it from the initial
> position of the point cloud?)
> #2. This threshold is measured in points? I mean, if I set the threshold to
> the value 100, this means that all the points that are 100 (points?) away
> from (each point/points) will not be considered?
> #3. As I will need to write a paragraph about this implementation, what
> authors should I mention?
>
> I will soon post my code here. It might help someone.
> Best regards,
> Nuno

If I recall the implementation correctly:

1) The idea of ICP is to A) find the closest point in set B to each
point in set A and then B) estimate a transformation using these pairs
of points. The "Too far away" means if a given point in set A does not
have a "reasonably close" neighbor in set B, we assume that this is
because there is no overlap in this region of the point set and
disregard this point in the computation of the transformation.

2) The threshold is measured in the same units as the data. I.e. if
the data points are stored in meters, the threshold should be
specified in meters.

3) I don't know if there is a specific paper to cite - I would just
cite the original ICP paper and mention this slight modification.

David



More information about the vtkusers mailing list