Faro: understanding the cloud to cloud registration settings

Discuss FARO SCENE software here.
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: Faro: understanding the cloud to cloud registration sett

Post by jcoco3 »

Yep, great points Phillip :) There is a great deal to try to explain, hopefully it will get easier in time. Great point about the "connecting scans" too! If you make sure you capture key scans that may help connect multiple loops then the registration should be better overall. I find myself adding these extra scans more and more.

A few questions popped into mind after reading you comments on subsampling:
1. Can anyone actually confirm the type of subsampling that is being used? From the use of "average subsampling point distance" you would think that a spatial filter is being used, but if not the effect that random decimation or row/column reduction would be dramatically different. My hunch says spatial filter, but I am uncertain it is that simple.

2. Also, do we know if scans of different resolutions are subsampled the same or if they are sampled proportionally to their original resolution? Maybe we should think about what part of the process is "averaged"?

3. With respect to your comment here:
I'd like to add that the density at which you had taken your scans, whether 1/4th, 1/5th, 1/8th, and the distance at which you've spaced your scan setups in the field, will highly change the effectiveness of subsampling. Since the density of points get less and less dense as you get farther from the scanner, if you've got lower res scans (i.e. 1/8th) you really want to keep more of your points, hence a lower subsample on lower res scans, to attain better registration results.
I completely agree with the logic here, but I have often wondered if a high subsampling is used during registration then are high and low resolution scans treated effectively the same with no perceivable difference in the registration results? Scanning at higher resolutions give you more opportunity to utilize lower subsampling, but I have found the point of diminishing returns approaches quickly for most environments. I think the only time it would make sense to scan at high resolutions is when you have very fine details that are visible across the project over great distances that need to be aligned very well. Alternatively I would think that scanning in a "perfect room" like I described earlier would produce the same results regardless of the resolution it was scanned, so scanning it at 1:1 would obviously be a waste if 1/8th or 1/16th resolutions registered nearly the same. Basically the scanned environment can dictate the C2C success more than the resolution, but they are both factors that need to be considered. I hope that made sense and my logic sound, but I am still uncertain.

Just as an fyi:
I just scanned half a dozen sites this week that I have scanned and registered in the past using C2C, and I must say there was far less effort in helping the registration process along this time. Yesterday, I registered (80) 1/8th res scans in just two steps, top down then cloud to cloud. No adjustments were made during or after. Last time I tried this I ran into significant problems that took multiple adjustments to resolve. It is very nice to see this kind of improovement in 5.5 :D Thanks again Faro Devs!
Scott
V.I.P Member
V.I.P Member
Posts: 1037
Joined: Tue Mar 29, 2011 7:39 pm
13
Full Name: Scott Page
Company Details: Scott Page Design- Architectural service
Company Position Title: Owner
Country: USA
Linkedin Profile: No
Location: Berkeley, CA USA
Has thanked: 205 times
Been thanked: 78 times
Contact:

Re: Faro: understanding the cloud to cloud registration sett

Post by Scott »

"Doubt is not a pleasant condition, but certainty is absurd." ~Voltaire

I greatly appreciate the extensive posts from Philip, Jonathan, and others concerning c2c registration and the 'black box' algorithms that make it all work (or not). Another recent LSF thread referenced the 'proof is in the pudding' concept that I also find useful, since that's ultimately where confidence in the process is reached. We all learn by doing and seeing what gives us the most consistent results from a variety of environments. I've never fully understood the math and physics that underly the software architecture (not even close), because it's hidden. I guess 'good enough' is the spectrum of understanding that I work within, and I'll just have to live with that. Thanks again for this forum.
JMontie
I have made 80-90 posts
I have made 80-90 posts
Posts: 80
Joined: Thu Jul 16, 2015 3:02 pm
8
Full Name: James Montie
Company Details: ReCon Engineering
Company Position Title: Piping Designer
Country: USA
Linkedin Profile: No
Been thanked: 4 times

Re: Faro: understanding the cloud to cloud registration sett

Post by JMontie »

IngSayyad wrote:I 'm just confused with the description of 'Sub-sampling quality slider: page 109 of manual.

see image:

I got higher mean value of 6.7 mm with high sub-sampling and spending less time than low sub-sampling which gave me mean value of 2 mm. But the manual says opposite or I just misunderstood the statement.
summary: C2C registartions
no of scans: 10
High sub-sampling: mean value of 6.7 mm (59 seconds)
low sub-sampling: mean value of 2 mm (2 min 42 seconds)
I'm still confused, the help says:
"Average point distance after subsampling of scans. For scans with large dimensions (like outdoor scans) a higher value is recommended."
So I understand that to mean the higher the setting the better the result will be, tighter tolerances. But you see from his numbers the higher setting generated lower tolerances in less time.

Jocco, your explanation may be right, but the help seems to say the oppisite. I ran into the same thing, I thought it was a glitch, the higher I bumped the subsampling the worse the tensions.

I think the problem is software engineers are writing the help files and validating the program. They understand the program and word the help files in a way that makes sense to them. Now novice user, using program and relying on the help files is lost. These are the people that need the help files the most.

In the military they had something called TM validation. They would take a Technical Manual and have entry level personnel follow the manual to preform the task in the manual. This caused changes to the manual, but it ensured the basic user could preform the task. This is how the help file should be structured, for the novice user not the software engineer.
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: Faro: understanding the cloud to cloud registration sett

Post by jcoco3 »

James, no need for any confusion Daniel pointed out they had an error in the manual earlier:
Hi Sayyad,

sorry for the confusion. You have found an error in the SCENE manual. I will forward it to the responsible person so it can be fixed.

To clarify the subsampling setting (for Top View and also Cloud to Cloud):
High - Will use less points for the registration. Faster but less accurate.
Low - Will use more points for the registration. Slower but more accurate.

Note: This is a rule of thumb and may not hold for small changes of the value.
Your military method for ensuring the quality of the help manual seems very appropriate in the situation. I think its a good idea :)
JMontie
I have made 80-90 posts
I have made 80-90 posts
Posts: 80
Joined: Thu Jul 16, 2015 3:02 pm
8
Full Name: James Montie
Company Details: ReCon Engineering
Company Position Title: Piping Designer
Country: USA
Linkedin Profile: No
Been thanked: 4 times

Re: Faro: understanding the cloud to cloud registration sett

Post by JMontie »

I reran my latest scan project with slider all the way to the low setting, It dropped the tensions from 0.0026m to 0.0003m.
User avatar
dan_mccurdy
V.I.P Member
V.I.P Member
Posts: 271
Joined: Wed Feb 15, 2012 2:34 am
12
Full Name: Daniel McCurdy
Company Details: Geometria Ltd.
Company Position Title: Technical Director
Country: New Zealand
Linkedin Profile: Yes
Location: Auckland, New Zealand
Contact:

Re: Faro: understanding the cloud to cloud registration sett

Post by dan_mccurdy »

It's worth noting that the high setting does produce better results when you're outside in variable environmental conditions (eg if it's windy or snowy). Making the subdivision too low seems to make it consider every small environmental change as relevant (which is great inside). An example we deal with regularly is scanning in snow and as you walk around your footprints change the snow level enough for the lower subdivision levels to freak out. The higher settings (ie greater space between used points) give a much better results, and sometimes are the difference between it working and not working. Even trees swaying in the wind too much can make it angry !
Dan McCurdy
Geometria Heritage Management
http://geometria.co.nz/
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: Faro: understanding the cloud to cloud registration sett

Post by jcoco3 »

I agree. I have tried to put what I think is happening into words here before, but to be honest I don't completely understand why it is better at the higher spacing. Very early on with Cloud to Cloud I had an experience with some very large trees in a very windy environment, I thought that there was no way C2C was going to work, but using less points(larger spacing) did the trick.

All I can figure is that by increasing the point spacing you are reducing the potential to have points of close proximity from scans that captured an object in motion, while retaining proximity for stationary surfaces. Since the object in motion represent no real stationary position and is more or less like random noise then increasing the point spacing increases the likelihood that there will be a greater positional difference between any points that might be compared. This should effectively give those areas less "weight" in the C2C analysis.

Sorry if this is confusing, I am just spewing thoughts in an effort to improve my own understanding :geek:
Hopefully this can evolve in to a clearer description of the process.
User avatar
dan_mccurdy
V.I.P Member
V.I.P Member
Posts: 271
Joined: Wed Feb 15, 2012 2:34 am
12
Full Name: Daniel McCurdy
Company Details: Geometria Ltd.
Company Position Title: Technical Director
Country: New Zealand
Linkedin Profile: Yes
Location: Auckland, New Zealand
Contact:

Re: Faro: understanding the cloud to cloud registration sett

Post by dan_mccurdy »

The way I have justified it in my mind ( and I don't know if this is correct), is using the analogy of creating a 3D surface over the points. If it is a wider spacing then the generalized surface won't include my footprints in the snow - it'll just be a broad, smooth snow bank, or the wildly waving tree (getting only the trunk). But if I drop the spacing down then the surface starts to include smaller and smaller features, like footprints or branches.

Thus it's easy to register a general snow bank or a tree trunk. But it's hard to register a very finely detailed snow bank with different footprints, or a tree with lots of waving branches.

I am sure there is a software engineer at Faro laughing quietly to themselves, with the actual algorithm on their screen, which no doubt involves hamsters and rapidly rotating wheels.
Dan McCurdy
Geometria Heritage Management
http://geometria.co.nz/
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: Faro: understanding the cloud to cloud registration sett

Post by jcoco3 »

:lol: :lol: :lol:

Your explanation makes sense to me, but I guess only the hamsters would know for sure :roll:
User avatar
Aaron_Hazen
V.I.P Member
V.I.P Member
Posts: 146
Joined: Wed Oct 29, 2014 6:31 pm
9
Full Name: Aaron J.R. Hazen
Company Details: Coast 2 Coast Surveys
Company Position Title: Senior Project Manager - 3d Architecture
Country: United States
Linkedin Profile: Yes
Has thanked: 1 time

Re: Faro: understanding the cloud to cloud registration sett

Post by Aaron_Hazen »

The higher setting should work better because it essentially increases the 'weight' that less dense points have on the calculation. It does this relatively. Raising the sub-sampling lowers the number of points across the board. However, more of the points deleted will be close to the scanner as that area was more dense in points.

Therefore, with more points being removed within a close proximity to the scanner and fewer points being removed at further proximity to the scanner, the further points are given more weight in the calculation.

Anyways, just came back to this post because I had forgotten the guide was wrong for 5.54 and needed to remind myself...
Aaron Hazen
Senior Project Manager

[email protected]
Image
http://imgur.com/a/CzfH4
Post Reply

Return to “FARO SCENE”