CloudCompare forum

Hi! I am trying to compare very "small" displacements between two big clouds (more than 500000 points each). I have overlapped them together by using Fine registration (ICP) aiming to minimize as much as possible the RMS. My best result was around RMS = 0.012 by taking 1000000 points and a tolerance of e-7. however, in my "point of interest" the movement given by the software after using Cloud/Cloud dist. tool is in the range of 0.008. How can this result be interpreted?
Thanks in advance!

Well, several factors can explain that, but 2 main ones are probably:
1) the fact that the RMS displayed by the ICP registration tool is computed only on the points used at each iteration (so 100k among the 500k you have). And if you use an overlap of less than 100%, this number will be even smaller. I believe the tool specifies the number of points it has used to compute the RMS.

A way to assess that is to compute the distances as you did, and then compute the RMS by fitting a Normal distribution on the resulting scalar field (with 'Edit > Scalar fields > Compute Stat Params'). The RMS will be displayed in the console (that's a hidden feature ;).

2) The C2C distance is exactly what its name says. It's the point to the closest distance (which may not be the right 'point', especially if the cloud density is low - see https://www.cloudcompare.org/doc/wiki/i ... omputation). You should use a more advanced comparison technique such as M3C2, especially if you try to detect small displacement (see https://www.cloudcompare.org/doc/wiki/i ... 2_(plugin)).