Cloud to cloud distances

[AEA]

Hi Daniel,
we have a real defoliated tree (trunk and branches) from which we have a very detailed 3D point cloud (not mesh) that we are using as a reference to assess the accuracy of a Mobile Terrestrial Laser Scanner (a system to scan trees). Once scanned, we want to compare the obtained point cloud with the reference. We are using the 'nearest neighbor distance' but do you think the local modelling would improve the comparison? According to the wiki, the quadratic model would be an option since we have branches and they are not straight. What is your opinion?

Many thanks,
Àlex

[daniel]

Yes, statistically it has been shown that the results are more reliable with the local model (because we only take the smallest distance between the nearest point, and the local model - so at least we can't find greater distances than the real one).

Regarding the choice of the model, definitely I would say that the Quadric surface is a good idea for a tree-like structure. However you need to select the local radius carefully so that the points neighborhood has a shape simple enough to be modeled thanks to a quadric. If the local cloud is too complicated / noisy, then the model will be useless.

[AEA]

Merci beaucoup,
We'll work on it.

Àlex