Sphere based registration is not accurate, I need some help

Discuss FARO hardware here.
Augusto 3D
V.I.P Member
V.I.P Member
Posts: 436
Joined: Sat Dec 06, 2014 5:26 am
9
Full Name: C-Augusto
Company Details: 3d scanning
Company Position Title: USA
Country: US
Has thanked: 115 times
Been thanked: 30 times

Re: Sphere based registration is not accurate, I need some help

Post by Augusto 3D »

So here is the crazy thing...
If I run c2c everything registers but the spheres seem "moved".

What does that tell you guys?
It tells me that maybe the spheres arent all that great quality wise regarless of what the vendor claims, or I dont know.
jorgen.olsson66
I have made <0 posts
I have made <0 posts
Posts: 1
Joined: Wed Sep 23, 2020 7:02 am
3
Full Name: Jörgen Olsson
Company Details: ATH Support AB
Company Position Title: CEO
Country: Sweden
Linkedin Profile: Yes

Re: Sphere based registration is not accurate, I need some help

Post by jorgen.olsson66 »

Hi Augusto,
Please check the "overlap" area in one of your scans. The area were the scanner start and end, the easiest way to check it is against a flat area like a wall on 20-40 meters. If this overlap is not even does it mean that the scanner measure wrong and this could explain your result.
I have a scanner in past which had that behavour and it was really a pain...
scankukai
V.I.P Member
V.I.P Member
Posts: 543
Joined: Wed Oct 05, 2011 10:14 am
12
Full Name: Scan ku kaï
Company Details: Scankukai
Company Position Title: CEO
Country: France
Linkedin Profile: Yes
Has thanked: 58 times
Been thanked: 207 times

Re: Sphere based registration is not accurate, I need some help

Post by scankukai »

Hello,

When I read the registration report, it seems that you have only 2 spheres between Exterior 001 and 003.

It also seems that Sphere and Sphere 2 in Exterior 001 create the highest tensions.

As Leandre wrote above, you should have 3 spheres between 2 scans, otherwise you have to rely on 2 spheres + inclinometer.
I don't know the distance you have between 2 stations, but the inclinometer can also create some tensions between scans.
So, one of the first thing to do when you have high tensions is to uncheck the inclinometer in some scans. But in your case, Exterior 001 and 003 won't be able to register correctly. In this case, try to create a plane in addition to the spheres.
Create 2 new clusters: the first with Exterior 001 only, and the second with all the other scans. Then, start the registration of the second cluster. If everything's fine, then you have a problem with Exterior 001.

So, when I see a report in which the same scan or the same spheres are always at the top of the tension list:
- I check that the sphere hasn't moved (as Lieven said, measure the distance between 2 spheres in 2 different scans),
- I uncheck the inclinometer of the scan, then I start the registration again,
- I isolate the scan in a new cluster and start the registration of the other scans again,
- if I don't have enough spheres between the scans, I try to create reliable planes.
Augusto 3D
V.I.P Member
V.I.P Member
Posts: 436
Joined: Sat Dec 06, 2014 5:26 am
9
Full Name: C-Augusto
Company Details: 3d scanning
Company Position Title: USA
Country: US
Has thanked: 115 times
Been thanked: 30 times

Re: Sphere based registration is not accurate, I need some help

Post by Augusto 3D »

yankoch wrote: Tue Jun 21, 2022 9:29 am Hello,

When I read the registration report, it seems that you have only 2 spheres between Exterior 001 and 003.

It also seems that Sphere and Sphere 2 in Exterior 001 create the highest tensions.

As Leandre wrote above, you should have 3 spheres between 2 scans, otherwise you have to rely on 2 spheres + inclinometer.
I don't know the distance you have between 2 stations, but the inclinometer can also create some tensions between scans.
So, one of the first thing to do when you have high tensions is to uncheck the inclinometer in some scans. But in your case, Exterior 001 and 003 won't be able to register correctly. In this case, try to create a plane in addition to the spheres.
Create 2 new clusters: the first with Exterior 001 only, and the second with all the other scans. Then, start the registration of the second cluster. If everything's fine, then you have a problem with Exterior 001.

So, when I see a report in which the same scan or the same spheres are always at the top of the tension list:
- I check that the sphere hasn't moved (as Lieven said, measure the distance between 2 spheres in 2 different scans),
- I uncheck the inclinometer of the scan, then I start the registration again,
- I isolate the scan in a new cluster and start the registration of the other scans again,
- if I don't have enough spheres between the scans, I try to create reliable planes.
thank you for your detailed response (and everyone else). I think that the culprit here was the positioning of the spheres and the fact that only a handful was used. I am going to run some more tests and try to do better planning next time. Maybe mix them with paper targets since the number of spheres is usually limited. Luckily this time I was able to fix the registration via C2C and using some planes as suggested.

From what I am gathering, it sounds like it is safe to mix spheres and checkers for target-based registration?
User avatar
landmeterbeuckx
V.I.P Member
V.I.P Member
Posts: 1616
Joined: Tue May 01, 2012 5:19 pm
11
Full Name: Lieven Beuckx
Company Details: Studiebureau Beuckx
Company Position Title: Owner
Country: Belgium
Linkedin Profile: Yes
Has thanked: 183 times
Been thanked: 548 times

Re: Sphere based registration is not accurate, I need some help

Post by landmeterbeuckx »

Augusto 3D wrote: Thu Jun 23, 2022 5:05 pm
yankoch wrote: Tue Jun 21, 2022 9:29 am Hello,

When I read the registration report, it seems that you have only 2 spheres between Exterior 001 and 003.

It also seems that Sphere and Sphere 2 in Exterior 001 create the highest tensions.

As Leandre wrote above, you should have 3 spheres between 2 scans, otherwise you have to rely on 2 spheres + inclinometer.
I don't know the distance you have between 2 stations, but the inclinometer can also create some tensions between scans.
So, one of the first thing to do when you have high tensions is to uncheck the inclinometer in some scans. But in your case, Exterior 001 and 003 won't be able to register correctly. In this case, try to create a plane in addition to the spheres.
Create 2 new clusters: the first with Exterior 001 only, and the second with all the other scans. Then, start the registration of the second cluster. If everything's fine, then you have a problem with Exterior 001.

So, when I see a report in which the same scan or the same spheres are always at the top of the tension list:
- I check that the sphere hasn't moved (as Lieven said, measure the distance between 2 spheres in 2 different scans),
- I uncheck the inclinometer of the scan, then I start the registration again,
- I isolate the scan in a new cluster and start the registration of the other scans again,
- if I don't have enough spheres between the scans, I try to create reliable planes.
thank you for your detailed response (and everyone else). I think that the culprit here was the positioning of the spheres and the fact that only a handful was used. I am going to run some more tests and try to do better planning next time. Maybe mix them with paper targets since the number of spheres is usually limited. Luckily this time I was able to fix the registration via C2C and using some planes as suggested.

From what I am gathering, it sounds like it is safe to mix spheres and checkers for target-based registration?
ALways try to have at least 5 in every scan.
Paper targets are imho only of use to connect to a coordinate system. You'll have to scan nearly perpendicular to them to avoid errors.
LSBbvba
Surveying services - 3D Laserscanning
Tel : +32477753126
www.lsbbvba.be
[email protected]
Augusto 3D
V.I.P Member
V.I.P Member
Posts: 436
Joined: Sat Dec 06, 2014 5:26 am
9
Full Name: C-Augusto
Company Details: 3d scanning
Company Position Title: USA
Country: US
Has thanked: 115 times
Been thanked: 30 times

Re: Sphere based registration is not accurate, I need some help

Post by Augusto 3D »

landmeterbeuckx wrote: Thu Jun 23, 2022 7:15 pm
Augusto 3D wrote: Thu Jun 23, 2022 5:05 pm
yankoch wrote: Tue Jun 21, 2022 9:29 am Hello,

When I read the registration report, it seems that you have only 2 spheres between Exterior 001 and 003.

It also seems that Sphere and Sphere 2 in Exterior 001 create the highest tensions.

As Leandre wrote above, you should have 3 spheres between 2 scans, otherwise you have to rely on 2 spheres + inclinometer.
I don't know the distance you have between 2 stations, but the inclinometer can also create some tensions between scans.
So, one of the first thing to do when you have high tensions is to uncheck the inclinometer in some scans. But in your case, Exterior 001 and 003 won't be able to register correctly. In this case, try to create a plane in addition to the spheres.
Create 2 new clusters: the first with Exterior 001 only, and the second with all the other scans. Then, start the registration of the second cluster. If everything's fine, then you have a problem with Exterior 001.

So, when I see a report in which the same scan or the same spheres are always at the top of the tension list:
- I check that the sphere hasn't moved (as Lieven said, measure the distance between 2 spheres in 2 different scans),
- I uncheck the inclinometer of the scan, then I start the registration again,
- I isolate the scan in a new cluster and start the registration of the other scans again,
- if I don't have enough spheres between the scans, I try to create reliable planes.
thank you for your detailed response (and everyone else). I think that the culprit here was the positioning of the spheres and the fact that only a handful was used. I am going to run some more tests and try to do better planning next time. Maybe mix them with paper targets since the number of spheres is usually limited. Luckily this time I was able to fix the registration via C2C and using some planes as suggested.

From what I am gathering, it sounds like it is safe to mix spheres and checkers for target-based registration?
ALways try to have at least 5 in every scan.
Paper targets are imho only of use to connect to a coordinate system. You'll have to scan nearly perpendicular to them to avoid errors.
That is true, but sometimes it hard to place balls everywhere - even if I had that many.
User avatar
Phill
V.I.P Member
V.I.P Member
Posts: 653
Joined: Wed Feb 06, 2008 7:19 am
16
Full Name: Phillip Nixon
Company Details: SKM
Company Position Title: Surveyor
Country: Australia
Location: Sydney
Has thanked: 7 times
Been thanked: 45 times

Re: Sphere based registration is not accurate, I need some help

Post by Phill »

Augusto

Check my post on this feed

https://www.laserscanningforum.com/foru ... 783#p86783

There is a report there. I did a whole stack of testing with spheres. Two things I noticed were - the laser will penetrate a small amount into the translucent objects (styrafoam, plastic) which is what a lot of these balls are made from. Basically if any light gets through it will give you an error similar to the one you are seeing as there is some amount of penetration into the object. I found that the only way to fix this was a layer of matt black paint then a layer of white over the top. The also changes depending on the laser wavelength (the C10 was different to the P40)

The other thing - not listed there but from experience when I use the spheres is quite often when I register with spheres the over spray at the edge of the sphere is included in the calculation and pushes the calculated target point back into the sphere slightly. I normally fix this by using a fence select and not selecting the edges of the spheres.

Saying that, I only use the spheres when I don't have access to my tilt and turns, and only for small amounts of scans. I also use Leica software and scanners (P40/Cyclone) so I cannot speak for Faro and Scene, but it might be worth investigating if these issues are affecting your target registration.

Hope this helps

Phill
Augusto 3D
V.I.P Member
V.I.P Member
Posts: 436
Joined: Sat Dec 06, 2014 5:26 am
9
Full Name: C-Augusto
Company Details: 3d scanning
Company Position Title: USA
Country: US
Has thanked: 115 times
Been thanked: 30 times

Re: Sphere based registration is not accurate, I need some help

Post by Augusto 3D »

Thanks for the info and links. I am using to Koppa sphere so I don't think that they are bad in this case.

We scanned another job this week and used a mix is spheres and self printed paper targets. I ran a quick registration and noticed that the most error was coming from 2 or 3 paper targets (0.5").

The spheres I believe were at 0.10 or so which is acceptable. (Less than 3mm). I am going to check the registration report again and post it here.

I did have to unassociate some spheres between scans and turn off the inclinometer (although I dont thing the later made a difference)

Speaking of home printed targets, I a matte paper better than a eggshells type?

I am going to order 2 or 3 seko tilt and turns.
User avatar
tadol
V.I.P Member
V.I.P Member
Posts: 126
Joined: Wed Dec 14, 2011 12:32 am
12
Full Name: Tad Laird
Company Details: Koppa Target Spheres
Company Position Title: Owner
Country: USA
Linkedin Profile: No
Location: Oakland, California, USA
Has thanked: 2 times
Been thanked: 4 times
Contact:

Re: Sphere based registration is not accurate, I need some help

Post by tadol »

I thought I would chime in, since the subject is about my targets -

I have not tested my targets for accuracy in a while - I sold a number to a large national laboratory known mostly for work on an atomic scale, and they did some careful testing with equipment far above my pay grade, and were impressed enough to buy a couple sets. I thought the numbers they got were way better than I'd ever hoped for. But that was some years ago -

So, with this concern being raised, I figured I'd justify my '46 Logan engine lathe and a little new tooling, and did a quick test - This was done with a random chosen 200mm target body, and a random base block. The bases are precision CNC turned in delrin, so I knew those were good, but the bodies are assembled from 2 halves, and there is a visible joint line where the halves meet, and this joint is sanded as a high edge is far more obvious than a low flat. So I have always assumed those areas would be a problem/and they are, but well within the tolerance of what a laser scanner is looking for. And those few points on this one latitude, if facing the scanner, when averaged with all the other points used to determine target center, would produce very minimal error.

Also, I would mention that my target bodies are made from a heavier and darker plastic that has almost no chance of the laser light penetrating or being absorbed, and we put a very heavy polyurethane paint on top of that - so, no light penetration is possible.

Here's a quickie video I took - the couple thou error in the base and a couple thou error on the circumference of the target could be partially explained by my 75+ year old lathe. But, the concentricity on the base, as determined by scribing the top of the body, is shown to be about as perfect as I could get - if the concentricity were off, the live center point would scribe a small circle -

I take great pride in making my targets as well as I possibly can, and am always looking for any improvement I can make and keep things reasonable in cost. If anyone has any questions or concerns, I'd be happy to address them -

Tad

https://youtu.be/DZLlghfWQC0
http://www.KoppaTargets.com
Inexpensive Laser Scanning Targets & Accessories
for Architecture, Engineering, & Forensics
"We're right on targets!"
Augusto 3D
V.I.P Member
V.I.P Member
Posts: 436
Joined: Sat Dec 06, 2014 5:26 am
9
Full Name: C-Augusto
Company Details: 3d scanning
Company Position Title: USA
Country: US
Has thanked: 115 times
Been thanked: 30 times

Re: Sphere based registration is not accurate, I need some help

Post by Augusto 3D »

No one is pointing fingers at your product, but just to be clear, are saying that the problem is not your spheres?

I personally can't belived that the majority of spheres that we put down seem to cause up to 0.5" of error many times. They are either moving slightly (so the magnet bases are sliding over time, some sphere are slightly deformed (the black shipping container doesnt really fit around them for instance) or our S+ series scanner is just not working correctly or all of the above.

You make a nice product and we are not putting down for the record. At this point maybe we need to invest twice the spheres or maybe faro scene is just not the "best software" when it comes to registration and many other things.

Everyone talks about having less tham 2.5mm of error using spheres.... we are using the same equipment and software and cannot replicate. Its not like are using a matterport or BLK360 or "ebay" low quality spheres...
Post Reply

Return to “FARO”