Scene- Manual Reg: Are more "Points" better?

Discuss FARO SCENE software here.
Post Reply
Superbob3D
I have made 30-40 posts
I have made 30-40 posts
Posts: 37
Joined: Sun Jul 05, 2020 4:03 pm
3
Full Name: Robert Howell
Company Details: Superbob
Company Position Title: Owner
Country: US
Linkedin Profile: Yes
Been thanked: 12 times

Scene- Manual Reg: Are more "Points" better?

Post by Superbob3D »

Hey All,

Wondering if anyone has a strong thought about this.

I'm registering a loop of 15 out door scans on a lot with dense vegetation around the perimeter. I did have sphere's set up with a min. of 2 spheres visible between scans, and there is one small structure on the lot to mark planes on. I've been tinkering with techniques to try to tighten things up registration and got wondering:

Is it better to limit the number of "Marks" (points or planes) to just enough for Scene to "Register and Verify", or are more marks are always better?

Thanks all!

b
User avatar
gord
I have made 20-30 posts
I have made 20-30 posts
Posts: 29
Joined: Tue Dec 22, 2020 9:30 pm
3
Full Name: Gord Anderson
Company Details: tiyuti
Company Position Title: developer
Country: Australia
Linkedin Profile: No
Has thanked: 39 times
Been thanked: 10 times

Re: Scene- Manual Reg: Are more "Points" better?

Post by gord »

From a purely math / geometry point of view.. I'd imagine that you'd want to have 3 reference points in common to match up any two scans,
[ with the 3 ref points not all in a line ]

If you only have 2 refs then you can match them up and still be out of alignment, by a rotation [ eg. the floor height might be different, so you could match up 2 level points and yet be rotated out of alignment slightly ].
gord anderson
quato.xyz
User avatar
Leandre Robitaille
V.I.P Member
V.I.P Member
Posts: 542
Joined: Sat Aug 03, 2019 1:53 am
4
Full Name: Leandre Robitaille
Company Details: Cima+
Company Position Title: Civil Technician - Surveyor
Country: Canada
Has thanked: 61 times
Been thanked: 246 times

Re: Scene- Manual Reg: Are more "Points" better?

Post by Leandre Robitaille »

I never use planes for final registration. Surfaces are just not flat enough for accurate use. Planes are only used to help position things if need be. Same goes for manual points. To much room for inacuracy. Also remember points are snapping to actual points captured by the scan. At 10m the scan does 6.1m density at 1/4. That gives room for +-6.1mm error just from that.

Always try to have 3 spheres between each scan at least. I often go for 4+
INV_Scanning
I have made 40-50 posts
I have made 40-50 posts
Posts: 46
Joined: Mon Nov 11, 2019 10:18 am
4
Full Name: Sebastian Winkler
Company Details: INVERS Industrievermessung und Systeme
Company Position Title: Project Engineer
Country: Germany
Linkedin Profile: Yes
Location: Essen, Germany
Has thanked: 5 times
Been thanked: 6 times

Re: Scene- Manual Reg: Are more "Points" better?

Post by INV_Scanning »

gord wrote: Mon Nov 29, 2021 10:51 pm From a purely math / geometry point of view.. I'd imagine that you'd want to have 3 reference points in common to match up any two scans,
[ with the 3 ref points not all in a line ]

If you only have 2 refs then you can match them up and still be out of alignment, by a rotation [ eg. the floor height might be different, so you could match up 2 level points and yet be rotated out of alignment slightly ].
It could work! If you put up the scanner nearly leveled and captured inclinometer data, the 3d problem becomes a 2d problem. So there is a solution!

All of that not taking care of a geometrically stable registration process. We normally work with:
- Enough targets (mostly spheres), at least 6 per Scan, distributed to all quadrtants of the scanner. Planes can help out. No clicked points.
- Only leveled scans (if possible)
- No open loops.
- Reference coordinates in difficult structures or larger areas.

Regards Sebastian
Emanuele
I have made 80-90 posts
I have made 80-90 posts
Posts: 85
Joined: Wed Apr 29, 2015 8:31 am
8
Full Name: Emanuele Ferrari
Company Details: Studio FERRARI
Company Position Title: Titolare
Country: Italia
Linkedin Profile: Yes
Location: Castellamonte (TO)
Has thanked: 9 times
Been thanked: 15 times

Re: Scene- Manual Reg: Are more "Points" better?

Post by Emanuele »

In open field situations I am using a topographic approach (forced centering) alternating scanners and spheres (of large diameter) on tripods.
I place a gps-rtk on the scanner, acquiring the coordinates for both verification of congruence and for any georeferencing of the survey.
User avatar
landmeterbeuckx
V.I.P Member
V.I.P Member
Posts: 1615
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: Scene- Manual Reg: Are more "Points" better?

Post by landmeterbeuckx »

If you want to do c2c in Scene in an urban setting you need to delete the trees or certainly the leaves before registration.

Maybe it works then....
LSBbvba
Surveying services - 3D Laserscanning
Tel : +32477753126
www.lsbbvba.be
[email protected]
User avatar
LPaulCook
V.I.P Member
V.I.P Member
Posts: 485
Joined: Sat Jul 18, 2015 9:42 pm
8
Full Name: L Paul Cook
Company Details: Land surveyor doing LiDAR
Company Position Title: President
Country: United States
Linkedin Profile: Yes
Location: Santa Barbara CA
Has thanked: 47 times
Been thanked: 44 times
Contact:

Re: Scene- Manual Reg: Are more "Points" better?

Post by LPaulCook »

If it's just one scan you may want to put that one scan in a separate cluster then both clusters into another cluster and register by top down cloud to cloud.

If it's more than one scan that only has 2 sphears then manually register and pick points rather than planes. As soon as the green bar at the upper right says register do it.

Sometimes you need to force correspondence using that tool on the left side of the manual registration screen.

When all scans are registered this way then do cloud to cloud a few times tightening the settings each time until you are satisfied.
L. Paul Cook, PLS
www.LPC3D.com
User avatar
jcoco3
Global Moderator
Global Moderator
Posts: 1724
Joined: Sun Mar 04, 2012 5:43 pm
12
Full Name: Jonathan Coco
Company Details: Consultant
Company Position Title: Owner
Country: USA
Linkedin Profile: No
Has thanked: 70 times
Been thanked: 157 times

Re: Scene- Manual Reg: Are more "Points" better?

Post by jcoco3 »

Please excuse my confusion, but I think some clarification is needed...were two different questions asked?

1.
Scene- Manual Reg: Are more "Points" better?
2.
Is it better to limit the number of "Marks" (points or planes) to just enough for Scene to "Register and Verify", or are more marks are always better?
At first when I read this I was thinking you (Robert) meant total number of scan-points or resolution, but then you went on to talk about the number of manually marked points. Number 1 could even be broken up into separate questions such as, are more scan points better for c2c or the verification in Scene, or do more scan points on the target surface such as a planar surface or even a target sphere influence its accuracy? If I am interpreting your post correctly I get the feeling like you are really only interested in answering number 2, and if so that would be good to know (please clarify) so we can focus on answering that question exclusively and leave out any answers related to number one.

If all we are looking at is question number 2 then it might still be a complicated answer. I generally tend to prefer what I perceive as better targeting and registration strategies as others have suggested, but I am not opposed to discussing the pros and cons of any registration method. So in your case, adding more manually marked points that are equally good or better could increase the registration accuracy, but where is the point of diminishing returns on registration accuracy for your needs with that particular set of scans? If you have a few good manually marked points and a disproportionately larger amount of poor quality marked points that are weighted the same, will this create a poorer registration solution?
When I say good or bad here I am referring to points that are clearly definable, positionally stable, and very similar between the different scans that observe that point. For example, the corner of some lettering on a black and white sign that is attached to concrete wall that is close and highly observable to most of your scans would constitute a better point than a more arbitrary asymmetrical discoloration on a tree branch that you did your best to mark the center of that is also very far from all scans that observe it and could have moved due to wind or other factors between scans or even during observation from an individual scan. Both are extremes of good or bad, but I hope you get the idea.

Sorry to ask more questions in the way I answered...just trying to help drive this to a clearer answer through discussion :)

This post made me think of...https://www.linkedin.com/today/author/d ... e-articles
I would bet the answer is in Daniel's Taming Errors, but I don't have time to re-read it in search of the answer today. Others here may be able to provide a clearer answer, so I hope they find this thread and chime in.
Superbob3D
I have made 30-40 posts
I have made 30-40 posts
Posts: 37
Joined: Sun Jul 05, 2020 4:03 pm
3
Full Name: Robert Howell
Company Details: Superbob
Company Position Title: Owner
Country: US
Linkedin Profile: Yes
Been thanked: 12 times

Re: Scene- Manual Reg: Are more "Points" better?

Post by Superbob3D »

Hello All!

Thank you all for the great replies! (And sorry for the late response.)

Super helpful insights. (This forum is AMAZING, ain't it?!)

Jonathan, you are correct that I was mostly curious to understand if there was, in fact, a diminishing return on adding extra "manually" tagged targets while registering.

I concur with everyone's attention given to setting as many high quality targets in the field as one can, and I am now understanding (as I suspected) that more "low quality", manually marked targets in Scene could actually be detrimental to the quality of the registration.

I have noticed that Scene will color code marked targets (ie, points and checkers), signifying (what I assume) is it's relative trustworthiness (i.e., Green is good, Yellow is OK, and it doesn't even let you tag something in the Red range). Does that sound correct? If so, I'm guessing you'd only want to use manual targets that were "in the green," if you have to at all.

I'm also hearing that, if, you are specifically tasked with scanning trees (and in addition to setting up as many spheres as one could) you would, ideally, want to remove the tree data prior to registration and put it back in after registration. If so, how does one eliminate data from a scan (for the sake of registration), but re-import those points back in after registration? Would it be the case that one could bring in a copy of a scan, delete out what you don't want for registration, then, after registration, apply the transformations (taken from the Properties info?) and apply it on to the original scan? Or is there a more straightforward way of doing this?

Thank you all again and I look forward to your thoughts!

Happy Holidays everyone!

Bob
User avatar
jcoco3
Global Moderator
Global Moderator
Posts: 1724
Joined: Sun Mar 04, 2012 5:43 pm
12
Full Name: Jonathan Coco
Company Details: Consultant
Company Position Title: Owner
Country: USA
Linkedin Profile: No
Has thanked: 70 times
Been thanked: 157 times

Re: Scene- Manual Reg: Are more "Points" better?

Post by jcoco3 »

This forum is AMAZING, ain't it?!
I wholeheartedly concur :D

I think you are on the right track, and most of what you said is how I perceive their system to work. This solution is generally not as simple as just more = better.

With regard to how Scene colors or classifies the targets...I generally think of this being something that a clever programmer did to make understanding the relative quality of a target fit simpler for the average user. By default it is scaled for most use cases in laser scanning, but some time ago they added the ability to change those thresholds to suit different needs which is great. I forget how many points must be acquired for Scene to label a sphere target as green...75-80ish I think. Obviously it is less for yellow or red, but we really should be asking ourselves, testing, and trying to understand what the quantitative effect of the number of points has on sphere fit and in turn the registration. In reality you can have zero to almost an infinite amount of points on target, so in the default scale of 0 to 80 points you have a trinary classification of red, yellow, green, but have you considered what it means if your green sphere has 200 or even 2000 points on it versus the default minimum of 80? The point that I am trying to make is that even red, yellow, and green is a simple way to look at what inherently has a much broader spectrum, and you may be able to achieve even better results if needed by going beyond the norm. Then again, where will that point of diminishing returns be? You can certainly over-sample a sphere without adding a benefit to its fit, but is a green sphere with 80 points truly equivalent to a green sphere with more points? I know this applies to more than just spheres, but it was the easiest example to use. I will admit, that I am unsure how manually picked reference points are effected by this, or what criteria dictate how their classification is calculated by Scene so maybe it doesn't even matter in your particular case. Perhaps somebody else here can shed some light on that.

For the trees...if your registration and verification is purely manual or targeted then the trees will have no effect on the registration so you can leave them in. If you suspect that the trees moved in the wind or are negatively effecting your cloud to cloud registration then you can remove them and then add them back in later in a number of ways. After registration you can manually copy the transformations from one project to another duplicate project(without the tree deletions), but there were a couple of plugins that made this task much simpler. A long time ago I had requested a feature where one could right-click on a scan and choose restore original scan data (minus transformations) similar to the restore grey feature. I am not sure whether or not that was every implemented. If it was my next request would be to make it something that could be applied at the cluster or global level rather than at the tedious individual scan level. Sorry I got off track there :lol:
Post Reply

Return to “FARO SCENE”