CloudCompare forum

Hi all,

I have been asked to take a LiDAR scan and produce a model. Please note that I did not know what LiDAR was a week ago.

Based on research I have done so far, I believe the steps I need to take start with creating Normals, then using the plugin PoissonRecon to create the mesh.

So I took this sample scan I was provided and used Segment to trip it down to just the chair & created Normals.


I then created the mesh.
The mesh is not defined at all and certainly not suitable to creating a model. I have changed a few settings trying to hit a sweet spot with no luck.

Any initial thoughts on what might be the issues? Any basic lessons on Point Clouds or Meshes you can recommend for a newbe?
BTW, I tried using MeshLab but cannot get anywhere with it and documentation is bare.

Any pointers / direction you can provide would be quite useful.

Thanks so much!!

Steven

So my main comment would be that your cloud is really noisy, and computing reliable normals on it is definitely a big challenge.

You say it's a 'scan'. What kind of sensor created this cloud? Do you have a grid structure maybe? (you would need to load it as an E57 or PCD file for instance). That would help a lot.

Otherwise other tools might have more options to create a surface mesh out of this... but I am not a mesh expert (hence the 'Cloud' in CloudCompare ;).

So how do you determine what "Noisy" is? How do I determine that?

I know it is a LiDAR scanner. Hand held I am sure. I have asked for the scanner info and will share it when I get it, but it is on a remote job right now and is incommunicado.

I will get an E57 and try again.

Thanks for the suggestions and information.

Steven

Sadly not all scanners will produce a cloud with a grid structure, and not all E57 files will embed this grid structure... but that's worth trying.

Make sure to use a recent version of CC as well (2.13.beta ideally).

Make sure to use the latest 2.13 beta version (10th of Feb) as there was a big bug with the E57 file export introduced on the 27th of January.