Problems with M3C2

[daniel_uca]

Hello,
First of all thanks a lot for creating this program! It is by far the most useful open source option I could find for my purpose.
I´m doing a study about the erosion of a coastal cliff by using photogrammetry. I read the paper about the M3C2 mechanism and it seems to be the most apropriate for my work. Nevertheless, I´m having some problems with it.

I´m still not very confident about which parameters to use in order to get good results but I think that I can maybe resolve my doubts by studying the paper more intensively again.

Anyway, after comparing two point clouds with the M3C2 mechanism I get as a result a point cloud that is completely black. Only when turning it in a particular orientation (frontal look between Z- and X-Axis) I can see the colors of the Scalar Field (see screenshots).




Is this some kind of error/bug or did I do anything wrong while introducing the parameters?

Furthermore, the presentation of the data when using M3C2 is by far not as demonstrative and useful as the result of a mesh to mesh comparison (see added screenshot of a fallen block).



Is there a way to make the result more attractive? Or should I just increase the number of core points to get a more complete representation of the cliff?

The fallen block from the screenshot is displayed in grey in the M3C2 result, even though the maximum depth of the cylinder should be large enough. Does this mean that the M3C2 mechanism can´t visualize this large scale change?

One last question: Is it possible to calculate the volume of surface change (i.e. the erosion) in either the M3C2 or C2M mode?
Kind regards,
Daniel

[daniel]

For the 'black' cloud, this is due to bad normals or a wrong orientation of normals. This happens when:
- the normal computation 'scale' is too small
- or the normal orientation heuristic is wrong

You can try different parameters with the 'Edit > Normals > Compute' tool to get good normals (instead of letting M3C2 compute them itself). With good normals, your cloud should look more like your mesh.

The color scale seems to be quite strange as well. Is it the default one? Maybe this is because of the bad normals...

Anyway you really need good normals to get proper results with M3C2 (as it's the base of the M3C2 distance computation). Have you used the 'Guess best params' button?

[daniel_uca]

Thanks for your reply!
It´s good to know that the error happens because of the normals. Yes I used the "Guess params" button but apparently the values weren´t very useful. I will try it again with different values for the normals.

I changed the color scale manually to get a better visibility of the changes so it´s not a problem with the computation but an intentional change.