Faster accurate terrestrial scanning with survey methodology.
We have a traverse survey methodology for terrestrial laser scanning, using total station, level and scanner observations, processed with least squares, for registration alignment, it provides very good results, especially in open areas where there are insufficient features for cloud based registration.
The scanner is a Topcon GLS-2000, capable of measuring prisms up to 200m, accuracy is what you’d expect from a 3 second total station. The scanner has a 1 second dual axis level compensator.
The scanner has a resection workflow, which allows you to move around freely, while targeting prisms mounted over control points, you can perform least squares adjustments on the resection, the vertical accuracy of the scanner compared a level isn’t as good as we need.
The level gives us sub mm elevation accuracy.
With our current workflow, there’s quite a bit of time required to setup tripods, put the scanner back in its case, move to the next location, leap frog the prisms, and measure side target control points. We measure instrument height in each position and come back and measure all the scan locations with the level. In each position we perform a low resolution full dome scan, then we take window scans of things we need to capture at higher resolution.
If we’re pressed for time, we can use only the scanner and level, without the total station.
It takes us about an hour for each scan position.
Having used scanners with VIS and IMU’s that assist in tracking the scanner location, these scanners can move through an environment much faster, with their light weight carbon fibre tripods, it only takes a few minutes for each position, however these scanners require sufficient features for registration and registration can take a long time if things go wrong.
I’ve been thinking about how I can speed up our scanning workflow.
The scanner itself only requires 1 minute and 45 seconds to perform a 12mm@10m full dome scan in greyscale. I’ve purchased a Ricoh Theta X, which I can mount on top of the scanner, it takes 360 degree 60MP HDR images in 10 seconds. Topcon makes a tubular compass, which I can use to point the scanner’s azimuth to magnetic north, which will help me orientate the photo’s later.
I’ve made contact with one of the carbon fibre tripod manufacturers, the tripod I’ve chosen is a hunting tripod, with a shoulder support carry pad, that makes it easy to move the tripod between scan positions, while keeping the scanner vertical. This is the tribrach support, underneath is a 25mm diameter Topcon type prism adapter.
This allows us to mount a prism underneath the scanner, so we can use the total station to track it’s position. We could use a 360 prism, however a nodal prism is more accurate and it’s easy enough to point it roughly at the total station, this is Sokkia’s AP12 60mm prism:
However we still want to level all our scan positions, the prism blocks off the scanners laser plummet, we could use an offset prism to give the plummet a clear path to ground, but the instrument height on the scanner is only accurate to 3mm, which isn’t good enough, so we normally use a Leica Height hook, but we don’t want to remove the scanner from the tribrach, and we don’t want to take angle instrument height measurements, the scanner has a trunnion axis mark, but it’s not really good enough either as the tape has to bend around the scanner to reach it.
So I designed something similar to a height hook which attaches to the AP12 prism, so we don’t have to remove it.
Continued…
- 45 Replies
The tripod:
A height tool, to measure instrument height, designed to accept the same Tajima 2m tape as Leica’s height hook, it has a 5/8-11 male thread on top, drilled with a 1/4″ female.
I was going to just attach a plumb bob, but I’ve found the laser plummet is faster for marking scan positions, this is necessary so we can come back later with the level. So it’s got a battery box for 3 x C cell batteries (8,000 mAh each). This laser projects a big ol’ bright cross on the ground.
I can use a tin template, with spray paint to mark each position quickly.
So the proposed new workflow is:
1. Set up nodal prisms over control points for resection on survey tripods, try to set these a good distance, so they don’t need to be adjusted too often, measure prism height on each.
2. Set up Total Station over a control point, to track the scanner.
3. Preload the scanner with a sequence of dummy points (so you can choose them, without having to type them in).
4. Level & setup the scanner tripod, point the scanner’s azimuth to magnetic north, take photograph.
5. Measure instrument height, add offset and enter it into scanner.
6. Select point position, start scan.
7. Measure scanner location with total station, while scanner is operational.
8. Turn scanner towards prism target, enter prism height and measure target (takes about 1 minute).
9. While scanner is measuring prism target, us spray marker and template to mark position on ground.
11. Move scanner to new position, rinse & repeat. Move the total station and control prisms as needed, pre plan locations to minimise the number of relocations.
12. Measure all scanned positions completed that day with the level.
13. Run obs from TS, scanner and level through least squares.
14. Upload the new point coordinates into the registration software and register scans.
I’ve been thinking about how this can be done with other brands of scanners as well. Other scanners would require spherical targets, as they can’t measure prisms. Not sure how to capture scanner observations for least squares analysis from other brands of scanners, maybe it’s not practical.
Anyway, just putting this here in case it’s of interest.
Cheers,
Peter.
Great stuff. I always love seeing innovative ways of matching up tools to accomplish the mission. Can I ask what it is you are doing that requires this level of precision and accuracy, particularly in the z axis?
Our needs/requirements are less stringent and we work on innovative combinations of drone, survey, gnss and ground based photogrammetry to get things done. I am always looking for new/alternative methods.
Maybe it’s not accurate enough for your needs, but have you looked at attaching a gnss receiver to the scanner and post processing it? A static-rapid static combination?
Thanks for the great post!
Does your Topcon GLS-2000 need to sight prisms? As opposed to being able to sight wall-targets? Just asking because we started getting better scan-scan registration when we scanned small windows around our targets with the finest scan-settings.
Also, out of curiosity, what Z-tolerance are you looking for? Are we talking 0.01’ at 30’ or are you trying to shoot for better than that?
We’re currently getting 3″ to 33″ on average, the highest reading, an outlier was 120″. Double these figures for 95% confidence. Topcon sell’s laser targets, but they’re not sphere’s or checkerboard targets like other laser scanners, they’re just a silver? circle on a paddle board, I don’t know their accuracy, in the manual it mentions prisms are more accurate. The prisms are measured separately to the scan, I suspect it’s doing so at a much higher resolution than the scans. What model scanner are you using?
This is from least squares, by default we apply a 3 second standard deviation, the STD Dev figures shown are over and above 3 seconds.
RTK is good if you can use it, I’ve designed a bracket for the top of the scanner, sometime in future I’ll get an opportunity try it. I post process static GPS when I need to reference the state grid.
The customer wanted sub 5mm absolute location accuracy, relative to their site monuments, and this was our second attempt, so we weren’t taking any chances, we used CloudCompare to filter outliers, the customer supplied us with GPS and plant coordinates of their monuments and levelling history since the 1970’s, we gave them local plant and state grid coordinates (no doubt the GPS coordinates will be outside their tolerance on the vertical). We improved the accuracy of their plant coordinates, after least squares analysis we went back to their original datums for origin and north, and found their height datum had been subsiding and chose a different monument that was historically stable. Their datums were stamped with their plant coordinates, we were able to reference these from their datums, so any surveyor in the field without the drawings will be able to figure it out. Originally the datums had been in a line of site of each other, but had been built out and were no longer visible to each other, we were able to align to their coordinates by processing our measurement obs with least squares.
First time around we engaged a subcontractor who had Faro and Trimble X7 scanners, these used a combination of checkboard targets and cloud to cloud registration, but the best results we got after registration was 35mm to 70mm location accuracy, we found out later that the Faro Scanners had calibration issues and this likely contributed (even if you have a recent calibration cert, scanners can get bumped in transport).
The environment had a number of large industrial sheds, with corrugated sheeting, we found out later that there were sufficient planes for Scantra to do cloud to cloud registration, had there been no calibration issues, however Faro Scene (the software used when scanning was performed) wasn’t able to handle the growing product stockpiles or the bags of product on pallets being moved around and unknown to us the calibration issues were causing problems too. We had problems with loop closure with scans, when you’d scan around the outside and inside the shed, the scans would be misaligned when you returned to your starting point. We also tried Vercator, they were very helpful, but ultimately couldn’t align the scans either (unknown to us we had calibration issues, but the software hadn’t identified it), I believe they’ve developed loop closure functionality since.
I was impressed that Scantra was able to identify the calibration issues, we sent about 200 of the scans for the Scantra rep to play with. I’m pretty sure that Scantra is the only point cloud registration software that can identify a calibration problem (happy to be proven wrong, by someone who knows otherwise).
Using the Topcon scanner with total station, level and least squares analysis, we’ll easily identify calibration issues, we didn’t know about Scantra at the time of our first attempt, but there are still areas where Scantra would require spherical targets, to tie scans together. The traverse workflow also allows scans to be spaced further apart as overlap isn’t required. This could be done with a scanning total station, but they’re much slower. We are considering using a Leica MS60 for scanning bolt connection details at distance, it can scan at 1 milligon point separation, where the best the Topcon GLS can do is 3mm@10m.
Registration of Faro and Trimble scans took months, required a powerful computer and ultimately failed, in comparison, registration with survey methodology can be completed in a day.
Cheers,
Peter.
Peter,
Thanks for the insightful reply. Sounds like interesting work. We get people wanting that kind of precision and accuracy, but not many that want to pay for what it takes to get it.
I’ll keep your name in mind if I come up against similar entanglements.
@gigharborsurveyor Thanks, if we can get the new methodology working it will bring down the cost somewhat, if we complete each scan position in under 10 minutes instead of an hour. We’ll still need to perform an initial survey, with the total station, level, static GPS and least squares to establish control points. The results will be more telling however.
@fugarewe Hi, I realised I misinterpreted your feet measurements for angles We needed to stay within 5mm absolute, 0.016 feet approx of all site monuments. The longest distance between two measured monuments on the site (with static GPS coordinates) was 1,745 feet. We did our best to accurately position and target backsights (for angular alignment), we used precise carriers, nodal prisms, adjusted all our level bubbles and tripod legs, even filed the tops of some tripods flat, before going to site. It was quite a challenge, but the results were very good, there were effluent tanks at Point C, the scanner managed to capture the tanks from a long distance (about 380m, 1250ft) from scan positions, that had no direct line of site between them, we couldn’t see any misalignment. These tanks weren’t part of the scope, we just happened to capture some points.
A couple more close ups.
Also looking at a foot assembly for working on grid mesh flooring typical in industrial facilities. Maybe a spike underneath? Normally tripod legs just go straight through gridmesh. We’ve used rubber mats previously, has anyone here got any ideas, or suggestions of things that work well in this situation?
Cheers,
Peter.
You just need to buy a Riegl Vz400i or vz600i and all these problems are solved at once. No targets, straight scanning without problems.
@beuckie I’ll get to try a vz600i in a few months, 30 seconds per scan I’m told.
@jitterboogie After something I can reposition quickly.
maybe a strong wheeled base with solid locking wheels like a hospital bed or industrial tool base.
@beuckie I’ll get to try a vz600i in a few months, 30 seconds per scan I’m told.
It is a beast but imho a vz400i is more rugged for outdoors. You can scan also at 30 seconds or 20 even to get excellent data outside.
I still haven’t had a chance to see our new(ish) vz400i in action personally but everyone who has loves it. ~30-45sec scans WITH imagery plus GNSS + IMU tracking…does everything but move itself to the next station automatically.
“…people will come to love their oppression, to adore the technologies that undo their capacities to think.” -Neil Postman
It is a beast but imho a vz400i is more rugged for outdoors. You can scan also at 30 seconds or 20 even to get excellent data outside.
What is the point spacing at 10m with those timings?
Cheers
@least I was told 6mm, which is all you want for full dome scans.
@beuckie Do you know if it’s possible to mark scan positions on the ground, does the vz have a laser plummet? We know our scan positions with a high level of confidence and mark them out, it’s easy for someone to set up a total station over a point later and use them when setting out. Some jobs are part of pre project planning in existing industrial plant, new installs and upgrades need to fit in with old plant, some things get removed, but other components remain, a lot of the existing plant gets obscures visibility when setting out.
Our std dev for elevations are sub mm when levelled, without the level they’re around 2mm:
The accuracy of the Topcon GLS laser plummet (3mm) is outside our tolerance, so we currently use a Leica height hook, but this is relatively slow using our current workflow and we also measure the marks using a total station, so we’ve got at least three instruments measuring each point location for least squares, it helps remove positioning errors, or identify any potential instrument calibration issues.
For our workflow we would still need to level a percentage of control scans. We did some tests with a Leica RTC360, marked chalk on the ground, close to each scan position and ran the level over it later and compared it to point clouds, this might be an alternative for the levelling workflow if we couldn’t measure instrument height.
Results we got from a RTC360, all scans within a 30m diameter cluster:
Point DL Elevation relative to SF 3002 RL Derived Scan Elevation m Scan Elevation mm Elevation based on differential levels from L6 Deviation mm Chalk immediately outside library entrance L1 1238.45 -938.04 50.343 -1.597 -1597 -1594.05 -2.95 Chalk on path, near library entrance L2 1305.4 -1004.99 50.276 -1.663 -1663 -1661 -2.00 Chalk on path, underneath enclosed pedestrian bridge L3 1216.8 -916.39 50.365 -1.573 -1573 -1572.4 -0.60 Chalk on path, beyond enclosed pedestrian bridge L4 1105.02 -804.61 50.476 -1.46 -1460 -1460.62 0.62 Chalk beside building B, beneath stairway landing 972.2 -671.79 50.609 0.00 Chalk outside entrance doors to building B L5 920.65 -620.24 50.661 -1.276 -1276 -1276.25 0.25 Chalk on path outside building B L6 933.4 -632.99 50.648 -1.289 -1289 -1289 0.00 Chalk on path, further from building B, loop close L7 1122.14 -821.73 50.459 -1.477 -1477 -1477.74 0.74 Survey Mark SF 3012 – screw in path near library 1434.3 -1133.89 50.147 50.168 Survey Mark SF 3002 – screw at end of path, near bollards 300.41 0 51.281 51.281 Regards,
Peter.
Log in to reply.