CloudCompare forum

Hi! I'm new here and have a few questions. I didn't find the answers when I searched. I'm doing research using CC's matching function. I still don't fully understand - I have STL files of a dental scanner (unit = mm). I want to get the difference in mm between the scans after superimposing. So far no problem. But do I get more accurate results by sampling Points on the mesh (both) before superimposing or would it even be less accurate?

Another question: How does it make a difference which STL file I choose as the reference and the aligned file? Theoretically it shouldn't make a difference?But unfortunately, I still get different results.

I look forward to your answers!

With recent version of CloudCompare, it's more accurate to keep the mesh (no need to sample points on it).

There's always a difference because distances computation is not a symmetrical (and the registration algorithm tries to minimize the distances from the 'aligned' cloud to the 'reference' one). In general, it's better to use the largest / most complete entity as reference.

Thank you for the quick reply - I've continued to work in the meantime. Now a new question has arisen. After matching, we work with the mean deviation and the standard deviation. I found the following formulas in the forum:

mean dist = sum ( distances ) / n

std deviation = sqrt ( sum ( Xi - Xm )^2 / n ) or
std deviation = sqrt ( sum ( Xi )^2 / n - mean^2 )

n: number of all distances
Xi: distance from Point i to its next neighbour
Xm: mean distance

My question is about the Xi. Said neighbor is the corresponding point on the other mesh? or on the same mesh? I'm not sure about the calculation.

Last question:
Is there a way to use only a part of a mesh for overlapping without clipping? We have intermediate areas that we want to exclude so as not to falsify the results. The rough overlap with point selection alone is too imprecise, since the mesh is fundamentally different (palate of an upper jaw).

Thanks!

I'm not sure which forum message it is, but anyway it's the standard formula to compute standard deviation from a set of values (here, it could be distances, but it could be anything). Xi is the ith distance value and Xm is the mean distance value. By default it would be the 'compared' cloud distance, but once again it all depends on which scalar field you apply this formula.

And regarding the question about 'overlapping without clipping', it all depends on what you want to achieve. If it's for distances, you could compute the distances on all the meshes, but then you'll have to clip/segment it to isolate the values (so it's better to do this in the opposite way, as the distances computation will be faster). And if you want to align the clouds with ICP for instance, then you don't have the choice, you have to segment the cloud first...