Normal dip direction seems reversed

[DBPRED]

New to cloud compare, so I apologize in advance for lack of understanding here. I have a SfM model of a boulder from which I have extracted two regions of interest. I want to be able to plot the normals of these surfaces to characterize surface coatings within the regions (e.g. lichen, and rock varnish). The idea is to do this with some sort of stereogram or rose diagram. To do this I have been computing normals, and then coverting to dip/dip direction. For a face of the rock which faces to the south, these normals appear correct (ca. 180 deg). For a face oriented to the north where lichen is present, the normals are also mostly oriented to the south as well, even though using the Compass > Plane tool appears to be showing the normals as pointing in the right direction. Could someone please offer some advice on what I am doing wrong here?

[daniel]

How did you compute the normals exactly? Because you have to take of how (per point) normals are oriented (that can be tricky). A point with the normal facing in the opposite direction will appear black.

And the 'Compass > Plane tool' may not work the same way / produce the same normals.

[DBPRED]

How did you compute the normals exactly? Because you have to take of how (per point) normals are oriented (that can be tricky). A point with the normal facing in the opposite direction will appear black.

And the 'Compass > Plane tool' may not work the same way / produce the same normals.

Thank you. Normals were computed with Quadric local surface model, an auto calculated Octree radius, and orientation using a minimum spanning tree of knn=6. I've tried several different settings, including using a preferred orientation on different axes, and can't seem to get it right. I've even tried to invert the normals with no success. Regardless of what I do, a majority of the normals on the face are facing in the exact opposite direction. They may display correctly on Cloud Compare, but when looking at the data in the stereogram, or as exported txt files in R, there are clearly a large number of reversed points. I'm wondering if I need to do a significant amount of cleaning of the data before hand?

[daniel]

Can you maybe send me the cloud (or a part of it)? (admin [at] cloudcompare.org)

[DBPRED]

Okay thanks for your help, just sent them your way.

[daniel]

Thanks for the data. See my reply.