Converting Point Cloud to Mesh in CloudCompare: Workflow and Best Practices for Different Environments

[Xifa]

Hello, I'm new to CloudCompare software and I'm trying to convert a point cloud into a mesh with good accuracy. I'm also working on creating a workflow for this process. So far, I understand that I need to clean the point cloud, compute accurate normals, and then create the mesh.

I would like to ask: Is there a standard process I can follow? Additionally, are there any differences in how I should process point clouds from different environments? For example, I have one point cloud of a room and another of a street. Are there any specific steps I should take when processing these two different types?

Also, should I convert the entire point cloud into a single mesh, or is it better to split the cloud into different parts, like the ground, walls, etc., and convert them into separate meshes?

[daniel]

Hi,

I am not aware of any particular process (apart the high level one you describe).

But clearly, depending on the environment / cloud, computing normals might be more or less challenging (or I should say more or even more challenging ;). I would say that the only way to ensure some quality here is to use structured data (Faro / E57 / etc. with scan grids, or at least with individual scans and the sensor position).

[Xifa]

Hi, I've seen the topic about scan grids and sensors discussed many times, especially regarding their effect on computing normals for point clouds. I know that sensors are shown in the point cloud as TLS/GBL in the 3D view. What exactly is a scan grid? Is it the yellow box? And what is the difference between using sensors versus using a scan grid?

[daniel]

The scan grids are only displayed in the cloud properties (as a list, as there can be multiple for a single cloud).

It's simply an array/table that gives the point acquired for each position of the scanner (generally organized as a regular grid because of the horizontal and vertical angular steps of the sensor).