Laser scan to mesh artifacts, why?

[C_Dordoni]

Hi All,
I am new to Reality Capture but have experience with Metashape for converting laser scan point clouds to meshes. There is an artifact present in Reality Capture that does not appear in Metashape. The source scans are in PTX format. I have meshed the same data set in both applications. Screenshots are attached.

Initially I thought this was just a display artifact in Reality Capture but when I render the mesh view I see that the artifacts are present in the mesh.

Why are there are drop outs or poor surface creation in the mesh in these trapezoidal areas?

My import and reconstruction settings are attached with some screenshots from both Reality Capture and the same areas in Metashape for comparison

Thanks for any feedback!

[OndrejTrhan (otrhan)]

Hi @C_Dordoni
This is quite strange, to be honest.
Can you try change the Point-cloud cropping radius to 150 and Minimal intensity to 0

and compute the model again?

[C_Dordoni]

Thanks for the suggestion, I’ll give that a try …

[C_Dordoni]

Here are some screenshots with the crop radius set to 150 and minimum intensity set to 0.

[C_Dordoni]

Here are some screenshots with the crop radius set to 150 and minimum intensity set to 0.

[OndrejTrhan (otrhan)]

Thank you for the test.
It looks like those pars are covered by something. How do look those LSPs (looking that way)? Is it possible that the original point cloud was cut?
Is it possible to share those data? If so, I will send you the invitation for the data upload.

[C_Dordoni]

Hi Otrhan,

There are no obstructions that would cause the dropouts. I am happy to share the PTX, so you may send me a link for uploading.
I would add that I have done some tests with using just a few of these scans as opposed to all of them and that seemed to improve the results somewhat.
Although I don’t need color and the scans are intensity only, it also seemed that setting the source to “Color” improved the result over setting to “Intensity”

I was (sort of) watching RAM usage (I have 64Gb) and I did not see it max out at any time. Based on some other apps I use, sometime just having more RAM available even if it is not used could make a difference.
Maybe I really need 128Gb if I am working with a larger # of scans.
There are 72 scans in this set.

Thanks for your assistance.

[OndrejTrhan (otrhan)]

OK, I sent you the invitation for the data upload to Box. It can be in your spam folder.

[C_Dordoni]

Got it, Thanks.

I am 7zipping the files now, I will upload shortly and let you know when its is uploaded.

[C_Dordoni]

files are uploaded now

[OndrejTrhan (otrhan)]

Hi, thank you for your data. I tested it and it seem this is related to the size of the reconstruction region. Have you changed it before model computation?
This is the result without changing the size of it:

and this is after set the bigger height:

and mesh looks filled:

[C_Dordoni]

Thanks Otrhan,

Yes, I changed it a bit.
So you are indicating that it is better to use the region that is determined automatically?

[OndrejTrhan (otrhan)]

No, the automatic one was bad, the modified region was better. In this case mostly the height of the region had to be changed.

[C_Dordoni]

OK, thanks for clarifying.