Alignment of point clouds

[justin_ro]

Good afternoon to all!

Currently I'm a graduate student trying to learn CloudCompare (beginner) and would need help regarding alignment of point clouds for my research. I need to look at the difference between two clouds and the change in elevation should be zero (or near zero) as these were taken a day apart.
After importing both, I manually align them with the Translate/Rotate tool, then run ICP (fine registration) with the reference being the "before" point cloud and the to be aligned being the "after" point cloud with RMS diff of 1.0E-6 and a random sampling limit of 10,000,000, while having rotation and translation with XYZ and Tx Ty Tz, respectively. Next is applying the Cloud/Cloud Dist tool (C2C) with max distance applied at 4.94 in this case, and having split X Y Z components applied.
This gives me a range of 0.00049m and 0.22m, which is a fairly large range. Furthermore, once done, these two point clouds are exported as DEMs (tools > projection > rasterize), to perform raster calculator in ArcGIS, where this gives me a mean of 0.0077m, which is fairly good, but not near zero enough.

My question is: is there a way to further reduce the range I'm getting? I would love to see a range of 0.0004m (0.4mm) and 0.001m (1mm) for example.

Any help will be greatly appreciated! Thank you in advance!

[daniel]

So pay attention when computing the distance in CC with the C2C distances computation tool, that the clouds density and non-overlapping parts (or holes) will have a big impact. It's better to use M3C2 for instance.

And for registration, it's a similar problem. Make sure the 'overlap' parameter is not set too high. Any mismatching parts will prevent the clouds from aligning properly. Also, a random sampling limit this high doesn't look to be useful.

Don't hesitate to share some screenshots, or even send us the clouds (admin@cloudcompare.com) if you want more support.