registration of complementary fragments

[LiseMC]

Hello,

I am working on a project of bone reconstruction.
I would like to re-assemble the surfaces of two broken fragments in the most appropriate position, then merge them.

After having manually re-assemble the point clouds, i tried to make an automatic registration in order to make the reconstruction more accurate.
It didn't worked , and replaced the moving cloud at a wrong place.
I tried to segment only the surface in contact, before ICP, but same result, I tried also to invert the normales.

What do I do wrong ? :-)

Thanks in advance for your answer.
Have a good day,

Lise

[daniel]

For the records, here's my answer (by email):

One important thing to consider when using ICP is that all points are considered by default, and the algorithm will try to match all points of the 'aligned' entity with the 'reference entity'. Moreover, if your aligned entity is a mesh, CC will use its vertices for the registration. It may be a good idea to sample (a lot of) points on the mesh first.

Then, to cope with the fact that by default all points are considered by the ICP algorithm by default you can:
- play with the 'overlap' parameter to give a hint to the algorithm that only a part of the aligned cloud should be used
- or clone the aligned entity, and segment it so that only the 'matching' part is used for the registration. Then you can apply the resulting translation matrix (in the Console) to the original entity.