Compute normals parameters

[boocaj]

I want to use the M3C2 pluggin to do some change analysis between two rockwalls in time. Instead of using the Normals parameters directly in the pluggin, I want to compute them before and then "Use cloud #1 normals" in M3C2. This should allow me to compute better normal considering there are more parameters.

My study site is however not planar. It is a rockwall with an almost 90 degree curve, so the eastern part of it is facing north, the middle part is facing west and the western part is facing north again (see last attach pictures).

I've tried different parameters (mainly for the prefered orientation) and obviously, Z+ and Z- aren't good because my rockwall is vertical, but Y+ seems better for the rockwalls facing north (Y axis) and X- seems better for the rockwall facing Ouest (X axis). As you can see in the attach pictures:

North facing rockwall with normals computed in Y+ are looking good, but west facing wall has some "shadows"

West facing rockwall with normals computed in X- are looking good, but north facing walls has some "shadows"

I feel like I could use the same "prefered orientation" (Y+ or X-) for the whole study site even tho both of them are better for some sections and worst for others. I am however wondering if I should segment my study site in 2 (north facing walls and west facing walls), apply the propper normals computation and then merge them back together?

Thanks in advance!

[daniel]

The only solution I see is to split the cloud in multiple parts (with the scissors tool), then compute/orient the normals differently for each section, and eventually merge all the parts together.

[boocaj]

Thanks for the quick reply!

I also tested the K nearest neighbours instead of specifying a prefered orientation and it seems to give good results for the entire study site (no matter if it is facing north or west). I however don't really find some detailed documentation on how to chose the good radius and the good KNN numbers.

For the radius, I can use the "auto" button to have an idea, but it gives different values every time I press it. Is it normal and how can I chose the right one?

For the KNN number, how do I "guess" the good value?

Thanks again!

[daniel]

That's the issue with this method of orienting normals, is that it's a lot of trial and error. There's no clear rule, as it depends a lot on the density of your cloud and also the noise...