This PR adds a symmetric version for transformation estimation that uses both source and destination normals and symmetrically splits the transformation being searched for over the source and destination points. It is an implementation of a paper recently published at SIGGRAPH19.
When used in conjunction with ICP it should have faster convergence and a wider convergence basin. To make this useful in cilantro we should add a new ICP method as well.
kzampog
commented
5 years ago
This PR adds a symmetric version for transformation estimation that uses both source and destination normals and symmetrically splits the transformation being searched for over the source and destination points. It is an implementation of a paper recently published at SIGGRAPH19.
When used in conjunction with ICP it should have faster convergence and a wider convergence basin. To make this useful in cilantro we should add a new ICP method as well.