CloudCompare forum

I'm currently doing basic distance calculation between two set of two merged scan and I'd like some advice. What are the main pro/con of choosing CloudvsCloud, CloudvsMesh or MeshvsMesh when it comes to registering/comparing 2 scans? I noticed that Meshes are far more timeconsuming but seems to give more accurate results?

Thanks you for your help!

It's all about the density of the reference cloud. If the density is high, then it may be sufficient to use the cloud directly (it's way faster than trying to mesh it). You can also use 'Local surface' (with C2C) or the more robust M3C2 plugin.

But if you have a good mesh, then you have absolutely no issue with the reference entity density and the distances can also be signed (well, M3C2 is also able to sign the distances).

See https://www.cloudcompare.org/doc/wiki/i ... omputation

Thank you! I managed to do several trials and it seems that either C2C-quadratic and C2M are the best for my project.

I got a new question so I'm updating this thread since it is similar.

When it comes to ICP, meshes takes more time to register but what are the pro and con of using C2C, C2M and M2M for registration? How can I compare them with my models to see what gives the minimum error?

I've read the ICP doc but it wasn't that precise :(

Thanks again for your help!

Well, meshes are slower to register this way, because the time to compute C2M distances is longer. But once again, if the clouds have a low density, then the result will be more accurate with a mesh as reference. For M2M registration, CC will actually sample points on the mesh automatically I believe (or use the mesh vertices, I can't remember). So there's no improvement here I would say.

My clouds is not that dense ( 16 point / unit^2) so I will work with meshes.

Thanks a lot!