Conflicting sphere target distances Faro Scene 2022.1

[lsf]

Happy Monday Everyone!

I have been struggling for a month on and off with this issue. A couple of people graciously shared their time and looked into it but with no luck. I did reach out to Faro support but there is no straight answer, and frankly it seems that this is not something where Faro sees the return on their time investment so I am left alone. For that reason I am reaching out to this community with a hope that someone else has performed a similar test and see what the findings were or perhaps there is some explanation for these results.

In the attached image you can see a compilation of 3 interior scans: scan1 - center position, scan 2 - right position, scan 3 - left position. I ran a command to find the targets automatically and I have assigned numbers to each sphere in each scan so make it easier to understand what is going on. I have underlined with blue color the distances that match within ~1mm between all the scans. You can see how all the distances match between the left and right scans but not with the center scan. I have calculated the variance between the conflicting distances and marked it with a black ink. I have highlighted the 'bad' distances with red color so that it is easier to visualize. I was hoping that maybe there is one or two spheres which are causing all of this, but I am not able to find the pattern. I get similar results with the external scans (center, left, right).

My instinct is that there is some issue with the target matching algorithm. There could be an argument that maybe the spheres are not perfectly spherical, but then why only the center scan fails. If every sphere would be oval in one direction, then I would expect to see the same issue in the left/right scans because the probability that I placed all the spheres in the same oval orientation in the center scans is very very low.

I am wondering if Autodesk Recap or Trimble Realworks would give different results? In https://www.dropbox.com/s/n5ks0cvdoilxt ... s.rar?dl=0 I have shared the 3 interior and 3 exterior scans. If there is someone who has the licenses and could send a screenshot of the same distances, that would be amazing. I don't expect anyone to have time for this, but if you have any idea of what could be causing this, please kindly share it.

EDIT: This website is reducing the quality of the image so you can find the source image here: https://www.dropbox.com/s/yld4h79inqixd ... s.png?dl=0

[TommyMaddox]

What is the actual diameter of the spheres you're using? And have you updated your sphere detection parameters accordingly? If not, you'll have mismatch as scene tries to fit a different sphere radius to the data set, leading to inconsistency errors.

I would also recommend manually marking sphere radius in selection tools as opposed to using the auto find tool, in close range the difference is not appreciable but it does move considerably as you increase the distance.

[lsf]

I have only one single sphere radius defined in the settings and it is 4" for 8" spheres therefore I think this should not be the problem, because left and right scan is okay.

[sreed]

I appreciate this analysis. Could you sketch up a diagram (simplified) that shows the problem/question? A diagram is easer to interpret, and perhaps you'll get more reviews/replied/suggestion. An alternative would be to show the correspondence top down (ortho) view to indicate the discrepancies.
Just a suggestion... as i'd enjoy seeing and experimenting with the same.

[lsf]

The Scan 1 (the left side of the image is the summary of everything, it shows all the targets, the distances and the discrepancies. Ignore the images on the right side.

EDIT: See the diagram attached, I hope that helps

[landmeterbeuckx]

The Scan 1 (the left side of the image is the summary of everything, it shows all the targets, the distances and the discrepancies. Ignore the images on the right side.

EDIT: See the diagram attached, I hope that helps

Can you share the raw scans please so i can import and check with Pointcab.

[lsf]

Yes, grab the link from the first post for the raw scans or click this url https://www.dropbox.com/s/n5ks0cvdoilxt ... s.rar?dl=0

[msivil]

I would not create the sphere automatically, if I would need to get accurate sphrere positioning. I would manually select the sphere surface areas and create the spheres.

[yankoch]

Hi,

There is also one thing to know: scanning points are less accurate at the bottom & zenith.
This is something I tested a few years ago, but it was with a Focus 120. I had differences up to 5mm.

I can see that you have spheres right after the bottom disk.

Try to place your spheres further.

However, a difference of 1-2mm is something common.

[lsf]

I would not create the sphere automatically, if I would need to get accurate sphrere positioning. I would manually select the sphere surface areas and create the spheres.

@msivil Thank you for your input! What software would you be using and how would you fit the inserted sphere into the right position, would you eyeball it? The greatest variance is 3.3mm. The Focus S70 that was used has +-1mm error out of the box (so 2mm combined) and I think that is for the objects closer to the 70m range. These sphere are practically next to the scanner or up to ~7m so I don't think that the hardware is the issue for this. Either way, I wonder whether manual sphere positioning would really be more precise and why.