Structured vs. unstructured file formats

[wftarch]

I'm just beginning to explore the world of 3D scanning and modeling and was hoping for some help with a particular issue. I'm working with a point cloud in .las format generated with a tripod based scanner, a Trimble model TX8. The scan is a "Level 3" according to the scanner, if that helps. The scan utilized target based registration with spheres. Someone else took the scan and processed the point cloud data using Trimble RealWorks. I've been experimenting with it in Recap Pro and CloudCompare. I would like to be able to generate a mesh from the point cloud, but understand from my limited research that it may not be possible in Recap Pro since the .las format data is considering "unstructured". But, I've also read that "unstructured" data is associated with scanners that don't stay in a fixed location while scanning, like those attached to a drone or a vehicle. So, in this particular case, since the scanner was fixed, can the .las file be considered "structured" or can it be exported or converted to a format that is?

According to Autodesk, an "unstructured" file format can't be converted to a "structured" format. From this, I assume that the "unstructured" file formats (like .las) just don't contain or support the data that makes for a "structured" scan, even if the scanning procedure was a "fixed origin" or tripod scan?

But, although I haven't gotten the results I would like, in CloudCompare I can compute normals and create a mesh from this "unstructured" point cloud.

Is the fact that CloudCompare can create a mesh from an "unstructured" point cloud file just evidence of the fact that it has more capabilities than Recap Pro? Are there any limitations for a mesh created from a .las file? With the right tweaking and fine tuning, will I be able to achieve the same results in quality and level of detail as I would if the mesh were created from a structured point cloud?

This is probably elementary stuff for most of you, but I would appreciate any advice, correction or clarification. Thanks.

[daniel]

Indeed, LAS files are unstructured because they have no information about the sensor structure, etc. It's really just a set of unorganized points (at least explicitly). If you have the choice, the E57 file format would be a good alternative (it can keep the structure of each scan).

Regarding mesh reconstruction, it's indeed easier and more robust to work from structured scans. This is probably why Recap relies on such an algorithm. However the Poisson Reconstruction approach is also very powerful, and has indeed the great advantage of requiring only normals instead of the scan structures. Of course it's not ultra-easy to compute nice normals, but depending on what your scans look like, it may not be impossible.

Note that CloudCompare is clearly not a meshing tool (contrarily to Meshlab for instance). But the Poisson Reconstruction plugin integrates an external library (developed by M. Kazdhan). And it is a very fine meshing algorithm ;)

[wftarch]

Thank you Daniel. You've helped me understand more fundamentally what I've been experiencing and what I can hope to achieve with the .las file.

[Oiz]

Additionally, if your scanner is a Trimble/FARO one, you can import the scans directly (if you installed the plugin for CC on Windows during the installation). This direct import preserves the structured "grid" as CC calls it.

Even if you don't have access or want to work with the processed data instead of the raw scans, there are two options:

1. Check if RealWorks allows to export/save the scans as .fls, .fws or .lsproj, as those are three Trimble/FARO proprietary formats that CC can open preserving the structured grid.
2. If you have a Trimble/FARO scanner, chances are that you also have the Scene software, the derpy brother of RealWorks. In there you can also process the data and export it as .fls, .fws and .lsproj.