Unstructured Scan To Mesh

[Buzz0m]

Hi,
I was wondering what the actual requirements are for the meshing to work. I know some free software require normal values (the direction from which the point was captured as the components of a normal vector) in order to do the meshing. Some can even estimate the normals for all the points in a point cloud without additional data. If one would be able to define the normals for an unstructured cloud, would recap be able to mesh the cloud? If so, how would the *.pts file need to be structured in order for recap to be able to mesh the cloud?

I'm posting this on a "solved" thread. Maybe somebody still reads these? 😃

[anil_mistry · Accepted solution]

Hi @Buzz0m,

Thank you for participating in our community! Project Clean up and Object Mesh only works with Structured scans.

Structured scan data is mainly coming from Faro, Leica, Z+F, Riegl, Topcon & Trimble 3D laser scanners in the format of .FLS, .PTG, PTX, ZFS, CLR, RSP and RDBX scan files respectively. These file formats, and some types of binary E57 scan files, are primarily structured scan data.

Structured scans are "fixed-origin" or tripod scans - data coming from scanners on a tripod (static or terrestrial scanners). For example, formats like FLS,ZFS,PTX etc. Only those files can be used for noise clean up and scan-to-mesh feature..LAS files that you are working with fall under unstructured. 

See this article for more information on structured vs unstructured format.

As far as I know, there is no software to convert unstructured data to structured.

Please, let me know if this helps.

Thank you and have a great day!!!

Please select the Accept as Solution button if my post solves your issue or answers your question.

[anil_mistry]

Hi @Buzz0m,

I'm just checking in again to see if you need more help with this. Did the suggestion I provided yesterday work for you?

If so, please click Accept as Solution on the posts that helped you so others in the community can find them easily.

[Buzz0m]

Yes you answered the question regarding the software functionality, but not what are the technical requirements for the meshing to work. For anybody reading this it seems that my original post, which was in another thread*, has been moved to it's own thread). The information you gave was partly given in the original thread and that post was accepted as a solution. Thank you anyway of elaborating the answer and taking the time to answer!

Maybe my question was unclear. What I was really tring to find out is that is there a way to "fake" a "fixed-origin" scan? By defining the origin the additional data given to a point is the normal (from which direction the point was captured). Does the recap meshing algorithm really need a specified tripod location or is the normal vector components really what is needed?

*https://forums.autodesk.com/t5/recap-forum/unstructured-scan-to-mesh/m-p/6441397/highlight/true#M636...

[ryan.frenz]

Hi @Buzz0m

Sorry but I can't share details of the meshing algorithm.  But in general, the service depends upon knowledge of not only the position of the scan, but also the validity of certain assumptions about the measurements and RGB/photo data within it.  Normals are part of that but one of the easier ones to be honest.  The service is still in Beta, and the limitation to structured scans is a simple way to ensure we can trust this info and provide a consistent experience.

You could probably fake an unstructured mesh by synthesizing fixed scans at intervals around the object.  You would need to make sure that the fake positions cover all of the data, and that each position only contains points which could actually have been observed by that position.  There is no support for this synthesis in ReCap.

Can you share more details about what you're trying to do?  E.g. what type of point cloud and object/scene?

-Ryan

[Buzz0m]

Hi @ryan.frenz,

thanks for taking the time to tell more about the meshing. The problem with the meshing is that my company gets clouds with different quality levels. You would be surprised of how inconsistent some of the data we get is. this means that getting a structured point cloud is not always a given because we sometimes even do manual file editing to clouds. And ofc we also deal with a lot of aerial scans, that I would like to include in the meshing. Our company deals with everything from indoor scans to large area city models so we're keenly following the development of recap and recap photo.

The approach where the algorithm uses normals seems easy and is fairly easy to understand and that's why I asked about it. If you've managed to mix in RGB colors etc that's awesome! Hopefully you'll develop the meshing so that these "additional" parameters aren't requirements but more like things that improve the meshing. IMHO everybody should understand, that meshing works with certain asumptions and is almost never a perfect depiction of what was scanned of photographed!

Keep up the good work and keep the community up to date on what you're working with! 😃 Do you have a newschannel or anything like that?

- F

PS. will there be a way of projecting points (from a referenced rcp-file) onto a section view in civil 3d in the foreseeable future? This is a much needed functionality. My personal main design area is underground spaces, so we have closed spaces that cannot be modeled as regular TIN-surfaces and hence there is no neat way in my knowledge of using pointclouds dynamically in sections.