Activity Feed › Discussion Forums › Photogrammetry, LiDAR & UAS › Laser Scanning – Best Practices
- Posted by: @pfirmst
@beuckie Interesting scanner, I haven’t had the opportunity to use one. Guessing you’d be using RTK with a base station, so c2c only has to rotate the clouds into alignment.
Rtk with only a rover antenna on it. C2c is done with gnss, compass and imu outdoors; imu and compass indoors. Goes blazing fast.
I do everything with it : houses, large facilities, roads, bridges,…..
We bought a Riegl terrestrial scanner early this year. I haven’t gotten a chance to put my hands on it yet since it’s been working halfway across the country, but the datasets I have seen come out of it are mind-blowing.
Riegl is also what we have on our top-end sUAS platform, and we get good enough ground returns under stands of trees that all we need are a couple of check shots rather than a full topo under canopy. Automated processing for flight lines, and the cloud from just the PPK+IMU trajectory matches levelled targets (checkpoints) within 1cm 95% of the time.
Amazing products. Expensive, but amazing.
“…people will come to love their oppression, to adore the technologies that undo their capacities to think.” -Neil Postman- Posted by: @rover83
We bought a Riegl terrestrial scanner early this year. I haven’t gotten a chance to put my hands on it yet since it’s been working halfway across the country, but the datasets I have seen come out of it are mind-blowing.
Riegl is also what we have on our top-end sUAS platform, and we get good enough ground returns under stands of trees that all we need are a couple of check shots rather than a full topo under canopy. Automated processing for flight lines, and the cloud from just the PPK+IMU trajectory matches levelled targets (checkpoints) within 1cm 95% of the time.
Amazing products. Expensive, but amazing.
+1 imho it is the best kit around. A bit heavy but after years of use i have my workflows and rolling carts.
Trimble uses Riegl sensors for their lidar products; they make the best sensors. I agree with all the above, the x7 is not the right instrument for this use. Take it back for an SX10/12. Pay attention to point densitity at set distances. If you are wanting one point per cm, then set your targets inside that distance. Keep set-ups simple, avoid 180 degrees, and remember Strength-of-Figures. Then your problems will all go away.
when I worked at Sanborn, we had recently purchased a riegl platform for one out new sensor systems. yeah. they’re the shizzzniz fo sho.
????
Glad I stumbled on this convo, really like reading thru everyone??s input.
We??ve had the SX12 for almost a year now and have had to come to terms with the very slow scan-speeds. It takes a minimum of ~25 minutes to scan a full circle at the coarse setting (and I think it takes like 45 minutes to scan a full dome at the coarse setting). That??s rarely something we can afford to do (especially since there will undoubtedly be a warning message indicating the level has changed since the beginning of the scan, with up to 0.05?? of error at 100?? or something). To make things more interesting, at distances greater than 40??, the coarse setting becomes frustrating to work with in TBC if you??re trying to be accurate (or trying to draft fast).
Our current workflow is to select small sectors at the ??Standard? point density (this minimizes scanning anything unnecessary, but it requires someone to be constantly tending the SX12).. and we also use two field personnel, one on the SX12 and another on a S5 (covering the same area). It??s totally redundant, but at least we come back to the office with traditional shots that we can run our automatic linework codes with, and whoever??s on the S5 is basically responsible for shooting things that would be frustrating or impossible to draft from the point cloud.
I guess I??m trying to say I??d probably avoid recommending the SX12 without a month or more of renting one to see if it??s limitations aren??t dealbreakers. As our dealer confessed, it shouldn??t be sold as a ??scanner.? It??s a total station that can scan?? except you can??t use it as a total station while you??re scanning?? and since it??s a slow scanner, you??ll be waiting around for a while before you can take any shots.
That said, I??ve been very curious if the SX12 could be enhanced substantially when paired with the X7/12. There??s almost no chance I could get my department to spring for another $100k+ of equipment/software/CPU??s/GPU??s, but I??m still curious.
Regarding @pfirmst??s comment:
I would make a small window scan of each target using a scanning total station, then you’ve got something for the X7 to register to..
Does anyone know if the office component of this procedure would be possible with TBC? (Or would you need Realworks or other software because of the larger point clouds from the X7?)
It’s both sad and frustrating to see firms buying equipment that works as intended, but trying to use it as something else…
If someone bought an SX12 expecting it to be “a scanner” in the same sense as traditional scanners, they didn’t do their homework. I think I can count on one hand the number of times that I have done a full dome (or even horizontal band) scan with the SX10/12 or MS60. Coarse scans are pretty rare too. I think I spend at least, if not more, time capturing images as I do scans.
I’d pair the S5 or the SX12 with a traditional scanner like the X7 or X12 (or another manufacturer if desired, we run Leicas and Riegls too), but putting them both on the same project is rarely going to boost efficiency unless there’s a time crunch or you have only one job to put people on.
Scanning total stations are best for those “in-between” projects that need more than a bunch of scattered single-point measurements, but less than 3D modelling. Properly used, they are faster than trying to pair a total station and traditional scanner.
Does anyone know if the office component of this procedure would be possible with TBC? (Or would you need Realworks or other software because of the larger point clouds from the X7?)
TBC works just fine with point clouds, even larger ones. I’ve pulled in aerial and sUAS data, clouds from Pix4D, other scanners, etc. They’ve been slowly improving the point cloud tools over the years. While TRW is the way to go for high-level modelling, on the survey side I just love being able to process and store all of my data in a single geodetic database…
I just recently did a locks scanning job pairing SX12 and X7 data, no problems at all except the X7 crew didn’t tie in the checkerboard targets with Perspective so I had to manually register everything. No problem there either.
I am hoping to give the X12 a test run, as nice and compact as the X7 is I still think the X12 will be more efficient for most of our work in the future.
“…people will come to love their oppression, to adore the technologies that undo their capacities to think.” -Neil PostmanThe idea that everyone else has the SX12 ??working as intended? has me thinking about starting a new thread, but I feel like the OP needs to be warned that anyone looking at obtaining this instrument will find literature that absolutely touts its scanning functionality – an understandable marketing strategy for Trimble – but only someone who is pretty well-versed in scanners will understand that the specs really don??t cut it in many use-cases.
With everyone here bringing up the SX12, I wouldn??t blame the OP if they thought it might make their life easier. I would love to hear the use-cases for the SX12 that people are having good experiences with (or specific examples for how to make the X7 work along the OP??s road corridor). Also, for what size budgets is this instrument really meant for?
I am not necessarily anti-SX12 – I wouldn??t necessarily want to get rid of ours – but in my experience, the SX12/TSC7 combo has been relatively buggy (not great for sending out with just any survey crew), and we are a small enough firm that the bundle was an expense we won??t repeat for some time.
In my experience, the following stats for the SX12 are pretty severe limitations:- a ??Coarse? scan over a sector, or slice of your total AOI usually takes about 10 minutes, but it will only make ONE pass.
- a ??Standard? scan over the same sector will take a little less than double the time, but it does multiple passes.
- the ??Fine? scan over any area larger than 5sqft is only acceptable if you can leave the instrument unattended while you accomplish some other task
- Regardless of your choice of Coarse, Standard or Fine, you will probably get a warning message about tilt error at the end of your scan – if you chose to use a Fine scan so you could do fewer setups, you may have noticeable variations between setups for areas far from the instrument.
- A setup with 5 Standard scans of reasonably squat sectors (scanning up to about horizontal from base of instrument), will take about an hour and will cover 360degrees (each setup will need a backsight/resection and check)
- If all your scans are done with the SX12, your setup locations will probably need to be spaced about 50?? apart (but opinions vary, @jitterbug is saying 150??-200?? at Coarse works.. not sure I??d want to have to have to draft with that density but hey)
- If everything went properly, you??ll get back to the office with cool scan data, in project coordinates.. and now the real work actually BEGINS (I hope you didn??t cut any corners because of how long the scans were taking).
Long story short, and for reasonable accuracies, you can probably cover 300?? of a 50?? corridor with a full day of scanning with the SX12 under ideal scanning conditions (add time for crowned roads, etc).
(and if you were hoping that the SX12 would keep you from playing frogger in the busy road.. remember you still need to have good geometry for each backsight/resection/check)
- Regardless of your choice of Coarse, Standard or Fine, you will probably get a warning message about tilt error at the end of your scan – if you chose to use a Fine scan so you could do fewer setups, you may have noticeable variations between setups for areas far from the instrument.
- If all your scans are done with the SX12, your setup locations will probably need to be spaced about 50?? apart (and each setup will need a backsight/resection, so add that time to each setup).
We never had a tilt error ever. might be the tripod or the surface you’re set up on too. I always used sand bags or CMU on each foot so nothing moved.
we also found through empirical data that our sweet spots for scans on coarse varied from 150 to 200 feet apart and never resected anything for c2c registration unless you count the primary control traverse and rounds which were independent of the scans anyway.
ymmv
I’m in Leica land now and deep diving into Infinity….
This is the solution I use in this situation, it’s a bloody good scanner, but there is a caveat, I must warn you that OEM support is not good and the processing software is some of the worst available. This is my workflow and it’s about as accurate as you can get provided your geolocation is also accurate. You also need to monitor air pressure, temperature and humidity and keep your equipment updated every hour or so.
First, I get a 1 second total station and differential level and mark locations where i want to scan, I set one of the points as my origin and create a local un-scaled coordinate system. Hint: Establish multiple base lines where possible and establish scan positions along them and make some redundant measurements between your baselines for least squares analysis.
The scanner I use is a Topcon GLS-2000, it uses prisms like a total station. I use a backsight prism about 100m away and occupy each point, measure a second prism position for checking the instrument accuracy in least squares analysis. If your backsight is too close you may lose some angular registration accuracy. Upload your point coordinates into the scanner using a csv file.
The scanner will be able to pick up prisms though mesh wire fences etc, you can’t do that with laser scan targets.
These are the scanning modes that I use:
1. Road mode – designed for roads, if you’re scanning a road, use this.
2. Standard mode – use this for long distance (500m).
3. Detail mode – use this for short distance scans (less than 100m).
4. High speed mode – I don’t use it, most of your time will be spent on setups.
5. Last Pulse – this will remove dust and allow you to scan through thin mesh fences if you need to. Also the laser frequency isn’t affected by mist, provided it isn’t too thick.
The software you want to use is Topcon ScanMaster version 3.0.7.4, it’s unsupported now, the registration module is made for surveyors and is easy to use, however you’ll need a license dongle. Try to limit the number of scans in each ScanMaster project to 20 scans, too many and it will crash. Magnet Collage has improved performance, however not all of the registration functionality was included, the traverse registration function wasn’t designed by a surveyor; it is confusing and difficult to navigate, if you have to edit values such as instrument height, it doesn’t behave as expected (from memory, my experience was the ground point moves vertically instead of the scan location).
Load the scans in ScanMaster, you can get the measurement observations (you have to calculate the backsight angle) from each prism and use them in your least squares analysis (this will give you feedback on the accuracy of the instrument).
After you do perform least squares analysis, upload your point coordinates using a CSV file into ScanMaster.
Register all your scans, then export them as PTX files, don’t try to generate normal vectors or use filters in ScanMaster. If you’ve done your measurements and least squares correctly, alignment will be perfect and is completed quickly, without manual adjustment.
PointCloudConverter from Kohera3D has a good mixed pixel filter in PointCloudConverter if you need one. The point clouds produced by this scanner are very clean.
CloudCompare can generate normal’s quickly from PTX files and scale your point cloud to your global coordinate system if you need to, it can’t save to PTX, but e57 or other formats work fine.
Translate the point clouds to their correct global coordinates if you need to.
In an open area this will be all you need to do to get good coverage.
Additionally, you can supplement your survey registered scans, when you’re in a area where there are lots of obstructions and you need many more scans for sufficient coverage, if there are multiple flat plane surfaces available for cloud to cloud registration. Corrugated sheeting is not suitable as an alignment surface, when I say flat, I mean brick walls, concrete walls, etc. If these aren’t available, or there are moving objects in your area (eg stockpiles growing in volume, vehicles, people and machinery), set out large diameter target spheres at random heights and distances, you don’t need to measure their locations, this allows them to be placed quickly. Use Trimble Real Works or Leica Cyclone to align free scans to your fixed survey registered scans.
The Topcon scanner can perform 12mm@10m full dome scans in just under 2 minutes per scan in detail mode, so it can be moved around quite quickly with a suitable tripod. Also it can perform small window scans 3mm@10m.
If you want to cover an area in less time use multiple scanners, you’ll be lucky to find two Topcon GLS-2000’s I’d suggest using other manufacturers.
Unfortunately the Topcon GLS-2000 becomes a single point of failure, if you have to send it to Japan, it won’t be back for many months.
It’s a great solution, but Topcon…
Don’t bother with the Topcon GTL-1000 or GTL-1200, read the laser scanner specs if there is any doubt.
@fugarewe Normal vectors are required by post processing software, eg to create a surface mesh (eg for water flow modelling), or to identify areas close to edges. If you want to classify your point clouds, these people are getting very good at doing that automatically, they will work with you to identify anything in the point cloud you want segmented, eg to identify gradebreaks. https://vercator.com/point-cloud-classification/
@pfirmst.. yowsa – I likey.. (I might try to do that ML myself) but let??s say the classification is complete.. then what? I thought I wanted points and linework.. a surface created from blessed points and breaklines.
I??ll try to work up the courage to start a new thread, but we are manually picking points so that designers can get a model that has specific points we have individually blessed..
@fugarewe After classification, break up the point cloud into smaller files for each classification, then when you import your points into a post processing application, you’re only dealing with the segments you’re interested in. Point clouds get very large, often too large to work with productively, by segmenting, there’s much less data to deal with in CAD etc. Also scanner measurement accuracy specs have a standard deviation, that means points may be up to 3 standard deviations from the mean, if std dev is 2mm, it means you could have a total distribution of 12mm, or 6mm either side of the actual surface, so you’ll also want to use a noise filter that removes points outside of the surface tolerance that you’re aiming to achieve. Then you’re ready to select points, create line work and surfaces.
- @pmfirst Eek.. I??m guessing you referred to this deviation in your previous post:
You also need to monitor air pressure, temperature and humidity and keep your equipment updated every hour or so.
Interesting. I like what you are saying, but I??m not sure I know what to do with those statistics (unless I just use them to report my level of accuracy).
I??m still in charge of reclassifying everything myself, so I think I??m not mentally prepared to deal with what you??re saying about the noise.
That said, I keep trying but I haven??t had much success with de-noising algorithms.. my current workflow basically stops after Cloth Simulation Filtering (where I mostly just play with the threshold, resolution and rigidness parameters).
I could see the noise-analysis being really useful for floor-flatness exhibits?? where else does it come up?
(If anyone else who??s reading this is struggling like I am, and you??re relying on CSF, you??ll still end up having to manually reclassify stuff you don??t want your clients to see (like bits of tires and cones, etc))
@pfirmst Regarding the breaking-up of the scan data into smaller files, I??ll give that a go.
I??m astonished I??m saying this but I??m guessing my first attempt will be to go through Recap (which finally has support for the Coordinate Reference Systems we use in our office, namely the current State Plane). I??ll see if it??s possible to draft on those smaller point clouds in Civil3D?? if not, I??ll try Infraworks.. if not Infraworks, I??ll try TBC. If nothing works I??ll just start swearing as quietly as possible and hope nobody realizes how much time I??ve wasted…
If nothing works I??ll just start swearing as quietly as possible and hope nobody realizes how much time I??ve wasted…
Oh yeah, laser scanning, when things go wrong, they go horribly wrong, try to get as many parts of the workflow automated if possible. I’ve found that using SALSA for least squares produces very good alignment results.
I also use a Leica height hook for instrument height, the Topcon scanner can measure instrument height itself, but only within 3mm, which isn’t really good enough.
Log in to reply.