Artefacts in deformation analysis

[Moja]

Deformation analysis of two epoch with three registrated scanner positions(for each epoch).

I get a result with cryptic artefacts like this:

https://imgur.com/gRdm8Rw

Any ideas?

[daniel]

Interesting pattern!

Are the coordinate very big maybe? (and you didn't apply the Global Shift?). Otherwise it may be due to artifacts in the scanning process, but it's generally not 'regular' so it's clearly confusing...

Can you share the clouds with me?

[Moja]

Thanks for your fast reply. The point size is default(small). Yes the global shift between both epochs is applied. Possibly a scaling problem (visualization)?

Here you can find the single database of the wall (.bin) illustrated in the image:
https://uni-bonn.sciebo.de/index.php/s/oay8SgTy6sOeJRr

... and here the whole dataset with some elements more. Shadowing can't explain the artefacts (specially at the top).
https://uni-bonn.sciebo.de/index.php/s/2DirO6qtxpCsPv4

[daniel]

Oops, sorry, it seems I waited too long before downloading the files. Can you send me new links? (here or at cloudcompare [at] danielgm.net)

[Moja]

database of the wall (.bin) :
https://uni-bonn.sciebo.de/index.php/s/gxOcuC2FLxXzaLJ

all:
https://uni-bonn.sciebo.de/index.php/s/yqLlaLNasHant9D

[daniel]

Thanks, I got them this time!

[daniel]

After several tests (C2C distance with or without local modelling, M3C2 distances, density computation, etc.) I didn't detect any 'algorithm' artifact.

The pattern has nothing in common with the octree (see image below), nor with the density pattern. And its magnitude is quite important (0.001 --> much bigger than the typical numerical computation accuracy which is around 1e-6 or 1e-7).

cc_artifact_against_octree.JPG (94.78 KiB) Viewed 2338 times

Therefore this pattern must clearly be already 'in the data'. Have any other pre-processing been applied to the clouds prior to the distance computation?

And how many scans are overlapped in this region?

[Moja]

Yes, the clouds are unified with cyclone.(maybe the reason for the artefacts)

8 scans from 4 positions, but the overlapping areas are different.(so i suppose that this is not the reason)