Point cloud accuracy loss after applying transformation matrix

[cragoras]

Hey, i just tried it at home with v2.12.4 and the same error occured. Doesnt seem like its fixed.

[daniel]

Could you send me a very small section of your original point cloud? (Just crop the cloud around your target for instance, and you can share it with me as a LAS file to admin[at]cloudcompare.org). Thanks.

[daniel]

Ok, as discussed by email, the issue was that you were expecting to apply the transformation matrix to the global coordinates, while it was applied by default to the local/shifted coordinates.

The latest 2.13.alpha version now offers a 'Apply to global coordinates' checkbox in the 'Apply Transformation' tool.

[Mehdimd]

Hello,

I am having the same issue, using the following transformation matrix:

0.96950086 0.26035895 -0.00356198 2540458.0100779
-0.26457433 0.96467088 -0.00097105 1154762.30123829
0.00318332 0.00188384 1.00413354 54.43521782
0. 0. 0. 1.

I tried to remove the translation part ([2540458.0100779 ; - 1154762.30123829 ; 54.43521782]) due to the huge numbers, and the accuracy was fine.
But I need the translation, therefore it is still an issue for me.

Any information regarding this? Thanks in advance

[daniel]

Have you tried this with the latest 2.13.alpha version? (on Windows only for now, but soon available with Flatpak as well on Linux)

[Mehdimd]

Yes I use the 24/06/2023 alpha version of CloudCompare on Windows. I have the same issue when just importing a point cloud with high value coordinates.

https://ibb.co/CVLP3jm

If this is not solvable, do you recommend any other workflow to be able to deal with those georeferenced point clouds in CloudCompare?

Thanks a lot for your reactivity