Unrolled cylinder not quite flat

[Brownie]

Hi there,

I tried to unroll a cylinder shaped tower but didn't quite succeed.
As is shown on the attached screenshots, the resulting cloud is still slightly curved.
Changing the radius only changed the scale but not the curve.
Deleting everything other than the core-cylinder did not change anything either.

Am I missing something? :-)

[daniel]

Well, several things can explain this, but I think the issue here is probably that the center is not perfectly detected. CC tries to guess it by computing the center of gravity of the point cloud, which is probably shifted in your case as the tower cloud is not complete.

In this case it's better to specify the center yourself.

To do this, you can:
- either pick two points (with the Point Picking tool or simply SHIFT+click on a point) opposite to each other and then compute the center of the segment manually (in the XY plane only)
- or add a cylinder to the scene (with the 'Primitive factory' in the file menu). Set approximately the right radius to this cylinder, and then move it around (with the 'Edit > Translate/Rotate' tool) so that it roughly fit the tower cloud. Then you can simply read the cylinder primitive center in its properties.

Don't hesitate to send me the cloud if you want me to look at this into details (especially if the issue was not related to the center ;).

[Brownie]

Hi daniel,

thank you very much for the fast response!
You are absolutely right. I figured it out a few hours later but couldn't edit the post because it wasn't online yet.

Since I'm new here I need to say it once: Your software is really awesome!
It is unbelievable how quick I was able to get results, thanks to the intuitive handling.
And that is saying something with such a complex matter. :-)

Anyway...

I tried once to find the best cylinder with RANSAC which worked fine but the unrolled result was still not what I was aiming for.
Would there be any chance of implementing a semi-automatic way to extract the relevant data from the created cylider?
Or move the center of the cylinder and any other selected entities to Zero (no need to transfer the coordinates) and - optionally - align the axis with Z.

The pickle is, that I am aming for a result, where all the features remain in the same vertical position as they are in reality.
Meaning that horizontal lines should remain horizontal in the unrolled cloud.
Is that gemoetrically possible at all?

I did get a result after unfolding the almost flat model, which was then not sliced apart by vertical lines (as is the case when doing thet with the original cloud).
However, the cloud shows significant deviations in horizontal measurements, sometimes 10% or more in varying directions.
Vertically, the measurements are absolutely correct.
Is that something I have to live with or is there any way of controlling the horizontal positions?

My guess is the deviations result from the varying distance to the base-cylinder for unrolling.
I tried to manually undistort the extracted raster image, but the result is not consistent either.
That is probably due to the fact that vertical lines in "reality" will not necessarily be vertical in the unrolled cloud.

Could it be done by projecting not vertically to the axis of the cylinder but parallel to XY?
Or would a distorted cylinder need to be the base for the unrolling, if that is possible at all?

Ok, that should be enough for now... :-)

[daniel]

Sorry for the delay, I was abroad (and I'm currently struggling to fly back home ;).

I'm a bit lost... if your cloud is properly 'aligned', the horizontal lines should remain horizontal after unrolling. And same thing for vertical lines.

And what distances are you referring to?

Otherwise, I agree that being able to use an existing cylinder entity to initialize the unrolling parameters would be great. I can add this to the TODO list, but I have no idea about when this feature will actually be implemented.

[Brownie]

No problem at all!
I hope you made it safely back home!
Were you trapped in the snowstorm in Northern America?

I only guessed about the vertical lines... :-)

What I mean is that when I unroll the cloud, the horizontal measurements do not at all reflect the measurements taken either on site or in CAD.
Verticals are perfect, horizontals way off.
Even when I adjust the resulting raster image to match the real measurements on the two extremities, in between the positions of features in the cloud vary greatly from their real positions - kinda like the cloud wasn't unrolled on a flat surface but a wavy one. Some features are further to the right as they should be, others too much to the left. Some are more or less correct, but that seems to be random. The magnitude of the deviation is 10% or more over relatively short distances (say 5-10 changes over the whole extent), although not visible to the naked eye.
I tried to visualize it on the attached image. The red arrows between the lines show the real positions an the calculated ones...

[daniel]

Isn't the cylinder axis off here? Or the center once again (even if it's less severe than before)

Otherwise it means that the actual shape of the building is not actually cylindrical and then you won't be able to get a clean result indeed.

Can you send me the cloud maybe? (cloudcompare [at] danielgm.net)

[Brownie]

Again, you are absolutely right.
I wanted to move the center to 0 but must have overlooked something.

My apologies!

Now, the deviations are only a couple percent, a few centimeters in reality.
And that I can live with, since some error is inevitable with such a procedure.
You might be interested to know that it DID work with the slightly leaning tower, although it is only 1-2 degrees.
Now the vertical AND horizontal measurements are ok.

Thanks again!