CloudCompare forum

Dear Daniel,
I am using M3C2 plugin applied to two lidar scans in order to quantify the geomorphological change of a moraine trough time. I read the article explaining how M3C2 works and I have understand that roughness is an important parameter influencing normals computation, and thus the real distance between the point clouds. During the field work I observed the grain composition of the moraine (cobble to debris size). Do you think that a normals scale of 10m, a cylinder projection scale of 0.5m and a max depth of 1m is good?

Thank you in advance.

Well, it also depends on the cloud density and scale of course. What are the scales of the objects and the global scale of the cloud? The normals scale seems a bit too much (but it's hard to tell). You may also want to use purely 'vertical' normals if you are looking for very smooth normals (well in this case it would ultra-smooth but also very fast to compute ;).

Point clouds have a scale of about 350x400m and each one resulted by the merge of four different scan positions. All the point clouds are composed by 61M points. Attached you can find the distribution of volume density (in m) and the roughness of a flat portion of the moraine.
I tried with the parameters mentioned above and the results seems pretty good...