would like some "behind the Scene(s)" insight on registration

Discuss all FARO SCENE related software issues here.
Post Reply
elina
I have made <0 posts
I have made <0 posts
Posts: 1
Joined: Wed Feb 17, 2021 11:00 am
Full Name: Elin
Company Details: I do not think it matters
Company Position Title: data specialist
Country: Estonia
Linkedin Profile: Yes

would like some "behind the Scene(s)" insight on registration

Post by elina »

So I have been reading this forum for some time, and have finally come to creating an account to answer all my questions I have not been able to figure out anywhere else (google, Faro Scene manual and webinars, that is).

The main question is - what would be the best practice/workflow/project structure for large multi-storey building? 
I am currently working clusters within clusters (floors/flats/rooms) in Faro Scene. We work with cloud-to-cloud registration, targets will be used just for georeferencing. One flat typically has 10-20 scans, but it still doesn't come together nicely - therefore I sometimes work with Auto clusters, sometimes create my own, but I haven't found a formula that works every time (even though floor plans are quite similar from floor to floor). It kind of troubles me, because for smaller projects I work with Recap and I know that Recap would have registered the same flats without problems and with much less manual tweaking.

I also sometimes have problems with registration precision - the report says everything is fine (1mm), bet in the correspondence view it is obvious that some walls are at least 1cm apart. I then try to bring everything closer with Visual registration (correspondence view), but it also doesn't work every time and I have to arrange/rearrange clusters.
I would also like some insight on Manual vs Visual registration - I like Visual much better, because we have floor plans and it just seems easier and faster. But could there be any difference in resulting registration precision?

How important in the registration algorithm is the first registration time? I am asking this because it seems that if the scans don't come together with Automatic registration the first time, it is later more challenging to get everything in place - compared to when I manually cluster the scans beforehand.

I have also noticed that, when entering automatic registration, pre-selected method seems to be different from time to time, even though I almost always choose the same one - is it some kind of software suggestion or has nothing to do with anything?

What kind of difference in your experience does subsampling value make in c2c registration? I don't seem to have noticed any traceable difference in results.

I have also had problems with project speed - when I have close to 1000 scans in one project, and >100 revisions, the project seems reeealy slow - sometimes opening project takes 30min, it is autosaving more often (or more noticeable), I notice some interface errors (for example, registration dashboard freezes, dragging and dropping clusters takes time).
Could number of revisions have anything to do with it? Should I remove some or maybe it is better to work with multiple projects and then merge them?
Is there any limit for number of scans in one project?
Our workstation is quite OK, therefore I doubt it could have much to do with hardware. 
How important is it with Scene to always have the latest GPU drivers updated?

Sorry, if this seems too many questions for one post:)
I do not consider myself a deeply educated professional, so please be gentle,
Elina
dmanning
I have made 30-40 posts
I have made 30-40 posts
Posts: 39
Joined: Wed May 20, 2015 1:44 pm
6
Full Name: Dustin Manning
Company Details: luxpoint
Company Position Title: Owner
Country: United States
Linkedin Profile: Yes
Has thanked: 8 times
Been thanked: 5 times

Re: would like some "behind the Scene(s)" insight on registration

Post by dmanning »

Here are my rules.

1. High overlap, whether cloud to cloud or targets.
2. Only use c2c when in a dead end situation, no loops.
3. always connect floor with targets.
4. scan stairways with very high overlap or targets. 1 scan at base, 1 half way up, 2 on each landing... etc..
5. Use good baming conventions, based on floor/level
6. Make hand drawn maps to keep yourself organized.
7. when in doubt take an extra scan
8. always be level
9. when using targets use the biggest spheres you can afford. this eliminates many movements in the field.
10. Undertstand that each room no matter how small will need 3 scans at a minimum.
INV_Scanning
I have made 30-40 posts
I have made 30-40 posts
Posts: 34
Joined: Mon Nov 11, 2019 10:18 am
2
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: 3 times
Been thanked: 3 times

Re: would like some "behind the Scene(s)" insight on registration

Post by INV_Scanning »

#2 is a really questionable statement in my eyes!

I would go the complete opposite way and call it: Do not use C2C with (especially long) dead ends. There is just no control in...
sim.herrod
V.I.P Member
V.I.P Member
Posts: 253
Joined: Thu Jan 14, 2016 12:26 pm
5
Full Name: Simeon Herrod
Company Details: Terra Measurement Limited
Company Position Title: 3D Data Manager
Country: UK
Linkedin Profile: No
Has thanked: 31 times
Been thanked: 53 times

Re: would like some "behind the Scene(s)" insight on registration

Post by sim.herrod »

dmanning wrote: Tue Mar 09, 2021 3:15 am 10. Undertstand that each room no matter how small will need 3 scans at a minimum.
Why? What is your thinking behind this?
Allen Nash
I have made <0 posts
I have made <0 posts
Posts: 1
Joined: Thu Oct 28, 2021 10:17 am
Full Name: Allen Nash
Company Details: Private
Company Position Title: 3D Survey Technician
Country: UK
Linkedin Profile: No
Has thanked: 1 time

Re: would like some "behind the Scene(s)" insight on registration

Post by Allen Nash »

Hello, Elin.

Apologies for the late reply.

In regards to your question about the subsampling, this was actually brought up in a post a few years back and was discussed quite extensively. Maybe you've seen it already.

viewtopic.php?t=9533

I found it to be very useful in my understanding of SCENE's C2C settings and have applied it to projects that I have worked on with good results. I hope that this helps you slightly.

Allen
Emanuele
I have made 40-50 posts
I have made 40-50 posts
Posts: 48
Joined: Wed Apr 29, 2015 8:31 am
6
Full Name: Emanuele Ferrari
Company Details: Studio FERRARI
Company Position Title: Titolare
Country: Italia
Linkedin Profile: Yes
Location: Castellamonte (TO)
Has thanked: 8 times
Been thanked: 5 times

Re: would like some "behind the Scene(s)" insight on registration

Post by Emanuele »

hi Elin, I try to answer based on my knowledge.

It is correct to create the clusters in cascade but you have to pay attention to the overlap between the scans that will be used to join them.
I create the first cluster with 3-4 scans registered with top view, then insert the others until the cluster is complete and if the registration is satisfactory I go to c2c; if the result is not good with the top view I align manually and then c2c.

There is a bit of mystery about the recording ratio and in fact even with low values ​​there are differences of 1-2 cm, however it must be considered that there are factors such as the level of the scanner, the noise, the size of the laser that affect on the final result.

The c2c recording is "statistical" therefore by repeating the operation different results are obtained.
I also noticed that it would seem affected by the first registration and for this I delete the scan manager with each new c2c registration.

The subsampling value is more appreciable in external scans or in any case of large objects; indoors, going from 3 to 5 cm changes little as there are still many points that Scene uses for recording.
Going from 1 to 10 cm can be seen some difference.

Scene "suffers" projects that are too big and with many revisions.
In these cases I make a backup copy and delete the revisions on the new one.
If the scans are also large, it is better to break up the project and merge it at the end with all the scans registered.
PatHughes
I have made 10-20 posts
I have made 10-20 posts
Posts: 12
Joined: Tue Feb 09, 2021 9:25 pm
Full Name: Patrick Hughes
Company Details: Position Partners
Company Position Title: Customer Success Manager
Country: Australia
Linkedin Profile: Yes
Been thanked: 2 times

Re: would like some "behind the Scene(s)" insight on registration

Post by PatHughes »

Hi Elina,

I think most of the issues you are facing with these large projects can be solved by using SCANTRA software to take in the FLS data and run the registration task for you.

The software can run a Plane2Plane type registration, this will fit planes within each point cloud and then calculate the registration parameters using this geometry, it is more optimised than trying to load billions of points for ICP registrations.

The software is built for tasks just like yours, where you have a large number of scans and the registration algorithms within Scene cannot produce a satisfactory result. Reach out to Daniel Wujanz on this forum or on LinkedIn and he will show you how to optimise your workflow so that you can save a lot of hassle and money.

https://www.technet-gmbh.com/en/products/scantra/
Post Reply

Return to “FARO SCENE”