Trust my survey or point cloud?
- gordonired
- V.I.P Member
- Posts: 144
- Joined: Thu Mar 09, 2017 4:27 pm
- 7
- Full Name: Gordon
- Company Details: Roska DBO
- Company Position Title: GIS Coordinator
- Country: Canada
- Linkedin Profile: Yes
- Location: Alberta
- Has thanked: 25 times
- Been thanked: 38 times
Trust my survey or point cloud?
Just when I start to feel comfortable with how all our projects are going, a wild curve ball has come my way.
We scanned a large gas production facility. We ran a survey control network, and had 4 high density/quality scans that had 4-5 surveyed points in each of them.
I'm starting with two survey controlled scans, and then doing C2C alignments of the connecting setups between those. I have reached a convenient "centre" point between these connections, and have found that those scans are wildly off from each other. The stack in the image shows how bad it is. On top of that, one stack has about a 1 degree tilt difference compared to the other. This has me leaning (pun intended) towards our survey being off, even though the overall error from the traverse on the total station was reported as being very low/good.
https://imgur.com/a/6ochwCq
Am I putting too much trust in C2C? I know there will be errors between two clouds aligned this way, but I'm not sure how greatly the error will propagate over the small number of scans (4), or the distance to that center point (~180 metres). We don't typically space our scans out 40 metres apart, I've just reduced the number of scans in this C2C alignment to take the most direct route.
We scanned a large gas production facility. We ran a survey control network, and had 4 high density/quality scans that had 4-5 surveyed points in each of them.
I'm starting with two survey controlled scans, and then doing C2C alignments of the connecting setups between those. I have reached a convenient "centre" point between these connections, and have found that those scans are wildly off from each other. The stack in the image shows how bad it is. On top of that, one stack has about a 1 degree tilt difference compared to the other. This has me leaning (pun intended) towards our survey being off, even though the overall error from the traverse on the total station was reported as being very low/good.
https://imgur.com/a/6ochwCq
Am I putting too much trust in C2C? I know there will be errors between two clouds aligned this way, but I'm not sure how greatly the error will propagate over the small number of scans (4), or the distance to that center point (~180 metres). We don't typically space our scans out 40 metres apart, I've just reduced the number of scans in this C2C alignment to take the most direct route.
- stevenramsey
- Global Moderator
- Posts: 1937
- Joined: Sun Aug 12, 2007 9:22 pm
- 16
- Full Name: Steven Ramsey
- Company Details: 4DMax
- Company Position Title: Technical Specialist Scanning
- Country: UK
- Skype Name: steven.ramsey
- Linkedin Profile: Yes
- Location: London
- Has thanked: 30 times
- Been thanked: 72 times
- Contact:
- sreed
- V.I.P Member
- Posts: 257
- Joined: Mon Apr 28, 2014 4:31 pm
- 9
- Full Name: Scott Reed
- Company Details: Prologue Systems
- Company Position Title: Principal
- Country: USA
- Skype Name: sabotva
- Linkedin Profile: Yes
- Location: Richmond, VA USA
- Has thanked: 16 times
- Been thanked: 15 times
Re: Trust my survey or point cloud?
I'm very interested to see how this thread develops. Our projects sometimes have the same conflicting constraints. A difference, however, is that we would not place as much faith in the C2C as we would the survey. I don't see a case where one would trust C2C over survey - especially with low residuals/closed network. There just isn't as much evidence or data to evaluate the C2C component of the registration, save for extreme slicing and dicing of the cloud.
The head scratcher for us is when targeted registration with low tensions (mean 2-3mm, max 6-7mm) fights survey control at project extents (15-20mm). In these cases we normally allow the registration to honor the TS control and "stretch" the targets used withing scan clusters. It may degrade the target tensions somewhat, but usually within a reasonable amount.
I'd be interested to hear others' perspectives on this curiosity and that offered by #gordonired.
The head scratcher for us is when targeted registration with low tensions (mean 2-3mm, max 6-7mm) fights survey control at project extents (15-20mm). In these cases we normally allow the registration to honor the TS control and "stretch" the targets used withing scan clusters. It may degrade the target tensions somewhat, but usually within a reasonable amount.
I'd be interested to hear others' perspectives on this curiosity and that offered by #gordonired.
- gordonired
- V.I.P Member
- Posts: 144
- Joined: Thu Mar 09, 2017 4:27 pm
- 7
- Full Name: Gordon
- Company Details: Roska DBO
- Company Position Title: GIS Coordinator
- Country: Canada
- Linkedin Profile: Yes
- Location: Alberta
- Has thanked: 25 times
- Been thanked: 38 times
- Formula1982
- V.I.P Member
- Posts: 162
- Joined: Tue Oct 13, 2015 4:02 pm
- 8
- Full Name: Stephen Ferrari
- Company Details: UTEC StarNet Geomatics
- Company Position Title: Technical Lead
- Country: United Kingdom
- Has thanked: 7 times
- Been thanked: 13 times
Re: Trust my survey or point cloud?
I'm not entirely sure how you've registered the data as it seems like a mix of total station control and C2C which is creating this mismatch.
Have you tried just tying each of the scans to the control using the targets only, and disabling all C2C, then checking for mismatch?
Have you tried the opposite - joining your 4 scans using only C2C, and checking for mismatch?
Finally, if the C2C only registration works well, have you tried registering the cloud containing those 4 (as a whole entity) into the control network?
There is clearly an error of some sort, but the above is what I would do to try and find where the error is. Whether it's a dud control point, whether it's poor C2C (eg if some of the elements in the scans have moved, if it's mistaking elements which are similar and near to each other for being the same surfaces, or if there is just too little overlap between scans), or whether one of the scans moved mid-scan and you didn't realise, meaning the actual 360 scan file contains dodgy data.
Have you tried just tying each of the scans to the control using the targets only, and disabling all C2C, then checking for mismatch?
Have you tried the opposite - joining your 4 scans using only C2C, and checking for mismatch?
Finally, if the C2C only registration works well, have you tried registering the cloud containing those 4 (as a whole entity) into the control network?
There is clearly an error of some sort, but the above is what I would do to try and find where the error is. Whether it's a dud control point, whether it's poor C2C (eg if some of the elements in the scans have moved, if it's mistaking elements which are similar and near to each other for being the same surfaces, or if there is just too little overlap between scans), or whether one of the scans moved mid-scan and you didn't realise, meaning the actual 360 scan file contains dodgy data.
- smacl
- Global Moderator
- Posts: 1409
- Joined: Tue Jan 25, 2011 5:12 pm
- 13
- Full Name: Shane MacLaughlin
- Company Details: Atlas Computers Ltd
- Company Position Title: Managing Director
- Country: Ireland
- Linkedin Profile: Yes
- Location: Ireland
- Has thanked: 627 times
- Been thanked: 657 times
- Contact:
Re: Trust my survey or point cloud?
Really difficult to say without asking a bunch of questions about your traverse;
- How much redundancy was there in your traverse? Number of rounds of angles, cross braces etc... ?
- Were you using tripods, bipods or single poles for your prisms, or wall mounted retros for resection?
- What method did you use to adjust it, e.g. 3d least squares, 2D+1D least squares, Bowditch or Compass rule and what were the closing errors like in plan, elevation and/or 3d?
- Were there any very short legs or other possible geometric issues in your traverse
- Did you bring in any external control, e.g. GPS stations, and if so, how was that computed? Does any external control include scale factor?
- What corrections did you apply, e.g. scale factor, atmospherics, refraction
- When have you last had your total station and scanner calibrated?
- gordonired
- V.I.P Member
- Posts: 144
- Joined: Thu Mar 09, 2017 4:27 pm
- 7
- Full Name: Gordon
- Company Details: Roska DBO
- Company Position Title: GIS Coordinator
- Country: Canada
- Linkedin Profile: Yes
- Location: Alberta
- Has thanked: 25 times
- Been thanked: 38 times
Re: Trust my survey or point cloud?
The control scans are too far away for any overlap at all. I need at least one C2C cloud next to each control scan for each to see any overlap. The overlap is nearly 200 metres away, so I was wary about even considering its accuracy . That being said, the overlap did look suspiciously off when I first tried this.Formula1982 wrote: ↑Tue Mar 10, 2020 2:26 pm]Have you tried just tying each of the scans to the control using the targets only, and disabling all C2C, then checking for mismatch?
This is what I'm working on right now. Scans are lining up as predicted, and the survey points seem to be a whopping 20 cm off at the one end when compared to scanned targets.Have you tried the opposite - joining your 4 scans using only C2C, and checking for mismatch?
The software refused to align the total cloud to the control network when I tried this. I had my constraints set at 20mm for each target/match. I could increase it, but that's fairly generous I believe.]Finally, if the C2C only registration works well, have you tried registering the cloud containing those 4 (as a whole entity) into the control network?
- stevenramsey
- Global Moderator
- Posts: 1937
- Joined: Sun Aug 12, 2007 9:22 pm
- 16
- Full Name: Steven Ramsey
- Company Details: 4DMax
- Company Position Title: Technical Specialist Scanning
- Country: UK
- Skype Name: steven.ramsey
- Linkedin Profile: Yes
- Location: London
- Has thanked: 30 times
- Been thanked: 72 times
- Contact:
Re: Trust my survey or point cloud?
And which software. your mixing scanners.
Is it really only 4 scans.
I usually find it is the settings and how you do the c2c that causes these errors as I do trust c2c when done well.
- Daniel Wujanz
- V.I.P Member
- Posts: 227
- Joined: Fri Aug 24, 2018 11:26 am
- 5
- Full Name: Daniel Wujanz
- Company Details: technet GmbH
- Company Position Title: 3D Laser Scanning Specialist
- Country: Germany
- Linkedin Profile: Yes
- Location: Berlin
- Has thanked: 206 times
- Been thanked: 188 times
- Contact:
Re: Trust my survey or point cloud?
Dear Gordon,
if you want I can have a look at your project. Error propagation, error detection based on registrations and other information is our bread and butter...
- poorly calibrated scanners
- no compensation when superior coordinates were used
- poor redundancy / bad network design
Cheers
Daniel
P.S. I'll cover the subject of quality assurance in "taming errors" soon. If anyone of you has a "head-scratcher" project, I'd be delighted to take use it as an example or at least to discuss it.
if you want I can have a look at your project. Error propagation, error detection based on registrations and other information is our bread and butter...
We typically review ~5 to 10 projects a week, where the top-3 reasons for discrepancies / tensions in the network are:..., but I'm not sure how greatly the error will propagate over the small number of scans (4), (~180 metres).
- poorly calibrated scanners
- no compensation when superior coordinates were used
- poor redundancy / bad network design
Cheers
Daniel
P.S. I'll cover the subject of quality assurance in "taming errors" soon. If anyone of you has a "head-scratcher" project, I'd be delighted to take use it as an example or at least to discuss it.
- stevenramsey
- Global Moderator
- Posts: 1937
- Joined: Sun Aug 12, 2007 9:22 pm
- 16
- Full Name: Steven Ramsey
- Company Details: 4DMax
- Company Position Title: Technical Specialist Scanning
- Country: UK
- Skype Name: steven.ramsey
- Linkedin Profile: Yes
- Location: London
- Has thanked: 30 times
- Been thanked: 72 times
- Contact:
Re: Trust my survey or point cloud?
These does look like a project crying out to be looked at and disected