Spatial Subsampling -SS not copying Image Pose Information

[LowIntensity]

Using Alpha 2.14 (Oct 14 2024)

When Spatial subsampling a point cloud, the image pose information is not copied (the images are), so the resultant exported scan images are not recognised in other software.

How to replicate:
1) import E57 with images
2) spatial subsample scan (any density)
3) export scan as .e57
4) reimport exported scan (into cloud compare or into E57 inspector)
5) compare the folder structure of the original (not SS) to the new (SS) scan
6) key differences are that the image pose information is missing (there might be other differences)

When comparing the two scans folder structure, there is no pose information copied for each of the cube face image. When then importing the output into the cyclone register, faro scene, e57 inspector or any engineering package i have tried so far, the E57 no longer imports with images in the right location with the scan.

The two attached images show the original scan folder structure with image pose information (Top) and the new folder structure of a scan, which was exported as e57 and then reimported into cloud compare (bottom).

If you need any other information, please let me know, I can also provide samples.

Thanks!

[daniel]

Thanks for the feedback. We will look into that.

[daniel]

Can you test the latest 2.14.alpha version?

I've updated it so that the image camera and pose information are always stored and transmitted to the subsets of points created during subsampling for instance.

Can you confirm that it works with other applications?

[LowIntensity]

Good Morning Daniel,

Great work it looks like you have fixed it. I have run a group of scans through the latest Alpha 2.14 and they have worked in a number of software packages perfectly.

I have a larger test running with a script overnight and will test a few more engineering packages tomorrow but things are looking great so far.

I appreciate the fast response, if i find anything else of note i will write back here.

[daniel]

Thanks for the feedback. Keep me updated if you face another issue.

[LowIntensity]

Good Morning Daniel,

i processed a number of scan projects (500 scans or more) and had no large issues with any of the new files. The new scans worked well with Leica and Faro software, which imported the images just as expected.

Reality Capture (a photogrammetry/scan meshing software) had a few issues, saying, "Warning: sampling estimation failed. Probably, the imported data are not a scan." but only for 12-14 scans of a given 500 imported scans. I am leaning more to the idea that given that all the other scans were fine its the implementation of reality captures e57 library that is causing this. however, I know that reality capture is very sensitive to null values or any values that contain special characters, so there may be some scope to look into that (if you need sample files for those few files that don't work, let me know, and I can send them through). Agisoft, had no issue with the same problematic scans and imported the images just fine.

E57 inspector interpreted the scans correctly too. I'll need to wait for spare licenses to test some of the other programs, but for now, I think there is a decent spread.

Thanks again for the fast turnaround

[daniel]

Thanks for the feedback.

Yes, don't hesitate to send a file that makes RealityCapture complain (I also have it on my side, this way I'll be able to do some tests). You can send it to admin@cloudcompare.org.