CloudCompare forum

Hello!
I'm new here, but I'm enjoying my time working with CC so far!

I'm a geographer that has been tasked with generating 3D plots of a salt brine cavern for use in a report. I have LAS files from a sonar caliper survey of the cavern that was done. I have examples of figures that were done in past reports showing 3D models of the caverns but we do not know what software was used to generate these 3D models. I've tried many different methods but so far I've come closest to replicating these results with CC.

Since I have no experience with this software, I'd like to be sure I'm creating meshes correctly. I'll link a google drive folder below with screen captures of what I've done so far. I also included the figures I'm trying to replicate as well as the LAS file.

First, I opened the LAS file with the default settings. Then selected the point cloud and computed normals.

Surface approximation: Plane
Neighbors: Octree, radius 17.217409 (clicked auto)
Orientation: Use Minimum Spanning Tree, knn = 50

Next, I used Poisson Surface Reconstruction with an Octree depth of 6 (this was the limit before weird artifacts began to show in the mesh).

My results are in the link below.

Two things that concern me:
The mesh does not fully make contact with all of the points in the point cloud (see concern 1 photo).
A "hook" was produced in the mesh. Can points be edited to remove this? Below that same area, mesh was triangulated even though no points exist there. I included photos of this (see concern 2 photos).

My apologies in advance for not knowing technical terms. I'd be grateful for any help offered. Thanks everyone!!!

Google Drive Folder Link: https://drive.google.com/drive/folders/ ... sp=sharing

- Will

Well, overall, not bad, you probably found the best route ;)

The Poisson Surface Reconstruction algorithm tries to fit a surface that pass as much as possible through all the points while still respecting the local normals and continuous surface. It's very hard, if not impossible, to produce sharp edges. And yes, the surface doesn't pass through all the points because it probably can't find a solution that respect the above constraints by doing so.

I think the biggest issue is with the cloud topology/organisation in the outer ring area: Looking at these points closely, I understand why the algorithm cannot find how to fit a single surface through all these points. Moreover, the estimation of the normals with so few points may not be super robust...

In your procedure, I would just change the 'normal orientation' method to 'Use preferred orientation' = 'From barycenter'. It's much faster, and is probably a little bit more accurate as it will consider the shape as a kind of sphere (instead of trying to deduce the local surface orientation based on the neighbors which are a bit messy in some areas). Apart from that very small improvement, it will be hard to avoid the various issues you spotted with such a cloud. You can maybe remove some of the points (with the Scissors tool) if you are able to understand which ones are causing these artifacts (especially the hook).

Is it possible to produce a surface with sharp edges, connecting all points of the point cloud? Are there options other than Poisson Surface Reconstruction?

There is no automatic and easy way to build the mesh for this difficult pointscloud
I tried Meshlab then auto-fill holes with Blender
You can manually use face creation in Blender to complete the mesh
later, you can merge or delete any point (hook points) with Blender
. . .

DA523 wrote: ↑Fri Dec 09, 2022 11:53 am There is no automatic and easy way to build the mesh for this difficult pointscloud
I tried Meshlab then auto-fill holes with Blender
You can manually use face creation in Blender to complete the mesh
later, you can merge or delete any point (hook points) with Blender
.a.jpg
.mesh (2).zip
.

Thank you for your response! This is exactly what I need to do.

Did you need to convert to different file type before importing to Blender?

las --> CC --> ply
Meshlab and Blender can import ply

DA523 wrote: ↑Sat Dec 10, 2022 12:18 am las --> CC --> ply
Meshlab and Blender can import ply

My apologies, I'm new to Blender. When I try to use the face tool or fill tool, I get the result shown in my attached images.

I've searched for advice on creating a mesh from vertices, but none have got me results like you show in your last screen shot. Could you offer any further tips?

Thank you so much.

Steps are
import stl in MESHLAB
filter menu
re meshing
surface reconstruction
export to stl
--------------
in BLENDER
import stl
change to edit mode
aa to select all
mesh --- clean up -- fill holes (sides =0)

For manual holes filling
click multiple vertices+ ctrl one by one ---mouse wright click -- new edge/face from vertices
check attached vid in zip file
..