Shifted pointcloud

[Manyyas]

Hello,

I have a little issue here. Im using surveyed point with coordinates to georefer the pointcloud. After transforming the pointcloud and aligning it to points, it asks me if I want to drop global transformation on reference points as well. Whats the difference please? I find this question/function as confusing one, I just want to have precisely georeferenced pointcloud according to surveyed points. Usually I choose NO - dont shift reference points. Once I found my pointcloud was like one meter off the position where it actually should be. So I did it again but I choose YES and it worked. But it doesnt make any sense to me.

I know the pointcloud is a bit shifted after import, but still dont understand the rest of it.
Can someone enlighten me please?

Thank You.

[daniel]

You (almost) always have to accept the shift, as it's a matter of accuracy (due to the internal representation). The applied shift will be stored by CloudCompare, and the original coordinates will be restored at saving time§.

Then in your particular case, I believe it's a matter of which 'shift' should be kept, as the 2 clouds may have originally large coordinates? So before overwriting the old 'shift' with the new one (obtained thanks to the Align tool), CC makes sure it's what you expect. But I understand it's not always easy to 'expect' anything in this case ;)

The only things that sounds surprising to me is that you said "if I want to drop global transformation on reference points". Reference points, if used as the alignment 'reference' entity, should never be modified.

[Manyyas]

Thanks for reply.

I know that input in large coordinates is in most of the softwares moved to smaller numbers. The thing was that after Align tool it asked something like: *Aligned cloud is shifted, but reference cloud is not. Do you want to drop global transformation as well?*
Maybe it told something a bit different, but it confused me. I thought it was asking just because of these internal computations, but it did matter if I have chosen YES or NO, because after exporting aligned cloud it was in different positions in either of these ways. Original pointcloud is usually in local system, and is shifted in import for example 0,5 meter. Do I understand it right, that maybe it takes this shift into consideration (depending if i choose yes or no) when exporting aligned cloud?

Thank you.

[daniel]

That would mean that you would indeed lose the original Global Shift when aligning to the 'reference' cloud which is not shifted. It's indeed quite unexpected: normally, the 'reference' cloud should have large coordinates (and it should be shifted), while the 'aligned' one would not.

In your case it seems that it's the opposite?

And do you really apply a shift of only 0.5 meters? (in which case you can definitely avoid the shift, it's useless)

[Manyyas]

Im not sure if it happens every time, 10 cm shift i would not notice, but half a meter is quite visible, if i compare it with surveyed lines. You are right that I can avoid shift...Surveyed point are in local system (just about x=1000m y=2000m), and scanner takes it in its own local system of course (about zero)...when importing .csv file its not asking me for shift of reference points, which I dont even expect to be shifted any time. But as I said, aligned cloud is shifted a bit after import and it seems to me, that it could export with this shift after georeference. That was my idea, because I tried it few times and nothing changed until I choose YES in that aligning question window.

[daniel]

Ok. I bet it's due to a limit case somewhere in your process.

If you ever have the time (and will), would you be able to share with me a pair of files that produce this scenario? (admin [at] cloudcompare.org)