Comparison of a cloud with a hollow model of steel structures

[clefdevoute]

Hello to everybody,

please, give me an advice. I am comparing a measured cloud (LiDAR, UAV photogrammetry) with a steel structure project (3D model). The aim is to demonstrate the overall accuracy of the methods against a reference project. Based on the histogram, I will present the total mean error, noting that it includes all previous mean errors (geodetic centering of inset points, cloud registration, etc.). I understand that the C2M method calculates the distance of a point from the nearest surface (mesh), which for "hollow" surfaces/models causes my problem. Whether I calculate distances without a sign (screen 1) or with a sign (screen 2), I cannot assign the reference surface to the points that represent it. I can manually measure several distances and calculate the mean error, but I would like to use the exact method of this program. Thank you very much

[daniel]

Maybe you could sample a lot of points on your model (with the 'normals' option checked) and then use M3C2 to compute signed distances? With M3C2, the corresponding points are searched along the normal, and you can choose how far from the source cloud they are, etc.

Not sure it will answer all your questions, but it would be probably cleaner / nicer. Of course, it's hard for CC to guess what's the right surface to choose for a point that's equidistant to two surfaces...

[clefdevoute]

Thank you Daniel for the advice and CC! I'll definitely try the plugin but I'm trying to find a way without manual intervention like sample selection..