CloudCompare forum

I have a question about the point-cloud to mesh registration in CloudCompare using the 'finely registers already aligned entities' option.

Does it use point-to-plane ICP? If not, how are meshes and point-clouds registered using point-to-point ICP? I found a reference to [1] as the details to the ICP implementation in CloudCompare, but [1] seems to deal with "two differently scaled 3-D point sets". How is this used for point-cloud to mesh registration? Is it possible that the vertices of the mesh are the only information used while discarding the face normals? Please help me out with this.

(Sorry if this information is already available somewhere else and I missed it.)

We use the the distance to the triangles actually (so almost as the distance to the plane, but constrained inside the triangle, otherwise we compute the distance to the edges).