Activity Feed › Discussion Forums › Photogrammetry, LiDAR & UAS › Interesting (and sort of funny) article on UAV accuracy checks
Interesting (and sort of funny) article on UAV accuracy checks
Posted by Dan Steely on June 8, 2021 at 11:38 pmncsudirtman replied 2 years, 9 months ago 9 Members · 20 Replies- 20 Replies
Checks should always be by an independent means/method and to the same or higher accuracy, whatever survey is being undertaken – otherwise they achieve nothing other than to waste time (and perhaps “earn” an hourly fee!)
We always carry out additional random RTK or total station checks on any UAV work we do, if only because there isn’t any other method of “calibrating” the results for different types of vegetation cover. A good sample of points within each vegetation type (ground vegetation obviously, not trees) provides a set of average corrections which can be applied to the area of each.
Software tends to like to average up the ground points along the boundaries between different types (or at hard edges) over a couple of metres width, so the check points need to be well within the surface type.
Yes, individual points will never be as accurate as a directly surveyed ground point, but with the corrections as above applied then the volume results will be as good as or better than ground survey. Anything critical, such as road edge heights can be surveyed as part of the check survey.
No, the method isn’t cheaper than a ground survey, IF DONE PROPERLY but it does give more information in a somewhat quicker time – and might be safer.
- Posted by: @chris-mills
No, the method isn’t cheaper than a ground survey, IF DONE PROPERLY but it does give more information in a somewhat quicker time – and might be safer.
And that is what 99% of clients, and about 98% of management types, do not understand.
Adding up all extras that a UAS survey entails, plus the required ground survey to perform proper QA/QC, means additional expense that won’t be overcome until the project is of sufficient size.
“…people will come to love their oppression, to adore the technologies that undo their capacities to think.” -Neil Postman Does the Pix4D have an IMU? AND is it tied to its own onboard GPS?
There has been discussions recently re the usefulness of classifying surface accuracy according to vegetation. bare, moderate ground cover, heavy ground cover, etc. Makes a lot of sense to me for remote sensing.
P4d is the processing software, not the drone.
@norman-oklahoma
I’m drone ignorant.
Its been dangled in front of me, and I’ve played with the toys versions.
I supported fixed wing OrthoPhoto and Lidar so got schooled heavily in that area.
@norm ASPRS came out with some accuracy standards that address that a few years ago. We follow these procedures for our large LIDAR projects. http://www.asprs.org/a/society/divisions/pad/Accuracy/Draft_ASPRS_Accuracy_Standards_for_Digital_Geospatial_Data_PE&RS.pdf
For sure – at an increased cost. Sometimes it is worthwhile, sometimes not. All depends on the project.
We have a remote sensing group in our geomatics department, and it’s pretty rare for them to be brought in without a clear objective or when we know for sure there are features that cannot be observed through conventional means, specifically due to the costs of doing it right.
Personally, I would love to fly all of our projects as SOP, just to have a current ortho to place behind the linework. But my organization is large enough that money and profit is literally the only consideration when deciding whether something should or should not be done…
“…people will come to love their oppression, to adore the technologies that undo their capacities to think.” -Neil Postmanjust curious what everybody’s thoughts are on the last paragraph of that article? statistically it makes sense of course but for general topo work where ground data needs to be within say +/- 0.05′ is that acceptable enough? Anybody have any experience using that method with photogrammetry? me personally I always like to pickup things either with GPS or conventionally with a gun where the vertical needs to be tighter than that for grade tie in purposes such as EOP, curb line, etc… as it is common place to have street grades with curb here that are at 0.5% and strip pavement where grades are 0.3%
I’m not a fan of doing hard surface topo with GPS. I use the total station for that. If my choice was between well controlled photogrammetry or RTK for hard surfaces, I’d probably go with the photogrammetry.
@ncsudirtman I believe the point is; if your error budget is +/-.05′, it does not matter what tool you use (who says my least squared hand tape triangulation network with a degree of freedom >1000000 couldn’t hit that), but rather can you prove that +/-.05′ independently? For example, do your RTK curb topo, build a DTM, then check that DTM with a digital level or total station and do some statistics on your independent checks. If you’re under .05′ @ 95%, then voila! In my experience, getting +/-.05 vertical @ a 95% confidence level is very difficult. That is usually my threshold for static control before the topo even begins. Let’s be clear, I am talking about network or absolute accuracy, not relative.
@dan-steely
Getting under 0.05′ vertical overall from a benchmark or datum can be difficult, but what I like about total station work is that relative to the previous shot 50 or 100 feet away, I am likely within 0.01′ or less, vertically. That is meaningful in design work and especially in staking. The biggest issue I have with RTK is that the real world expected precision is perhaps 0.1’…PLUS OR MINUS…meaning that two shots 5 feet away, I am likely to have one down 0.05′ and the other up 0.07’…and when you add that up, it can make for ugly curbs.
-All thoughts my own, except my typos and when I am wrong.And to follow up…if you had a large site, and you wanted to be ACCURATE to a certain benchmark, RTK might be exactly the tool for the job. The precision shot to shot might be less, but it that precision and accuracy is maintained over a much larger area than a total station (that isn’t moved).
-All thoughts my own, except my typos and when I am wrong.The biggest issue I have with RTK is that the real world expected precision is perhaps 0.1’…PLUS OR MINUS…meaning that two shots 5 feet away, I am likely to have one down 0.05′ and the other up 0.07’…and when you add that up, it can make for ugly curbs.
Yeah, but that’s why nobody grades anything with GPS.
@bstrand – so I agree and I don’t agree with you as I work alongside a lot of grading contractors. And in their defense the automatics of GPS on equipment has come a long way relative to 2 decades ago (or so the salesman tells them) but I don’t like using it to fine grade with personally if the model isn’t perfect – though I have had a contractor demonstrate his base & receiver along with a good model were capable to within 0.02-0.03′ vertically in situations where signal was good. Still I am weary to trust GPS for some of the most crucial things like laying utilities at minimum grade (heck I don’t trust a laser for pipe really when it’s flatter than 0.5%).
The good contractors know when to bust out a level or at the very least a laser & check stuff that might look funny to them. The bad ones will blindly use the GPS (and sometimes knowingly) blame the designer/engineer or the model builder when it doesn’t work out only to try to change order for it later. Had a 2nd hand story the other day about a very, very large grading & utility outfit that let the site super instruct the grade foreman and his crew to install the handicap ramps wrong (per the GPS model) and they’d change order for it later for tearing them out & replacing them just to knock another item off the punch list. The 4′ digital level clearly showed things were off per ADA as well as with the HC parking spaces but they paved & poured concrete for the ramps anyway because the GPS told them to do so. Building inspector noticed it & they still didn’t get the CO for the building but they claimed they’d completed their portion of the punch list that the GC had. That’s very wrong IMO but such is life & hopefully the client learned a hard lesson about those who over-promise at the beginning of a project & under deliver at the end as there’s something to be said about those who throw up red flags at the beginning & along the way. apologies on the very long tangent haha!
Yeah, I should have said grade anything vertically important with GPS. heh For certain things sure it’s totally fine.
We have different techniques and equipment for a reason.
If it’s practical (and the site is big enough) the UAV will give a good general accuracy (0.1′ or so) with a picture (so you can find all those things you would have otherwise missed. The GPS will provide a bit higher accuracy onto things you are going to build up to, but not join – such as the general road levels around the site or the watercoiurse bed levels. The total station will provide highest accuracies to ensure road junctions are smooth and even.
It’s a case of time versus accuracy: on a large enough project (or difficult enough site) then all three methods have their place – use what is best suited (or what you understand!)
@dmyhill – agree there completely. I see way too many these days who just trust the automatics of GPS for fine grading streets or other subgrade crucial surfaces with newer dozers outfit with automatics that are advertised as being capable of such. Then the paving crew shows up to complain about the uneven grade rocking the paver’s screed or somebody will say the pavement thickness in the asphalt reports wasn’t to spec when in my opinion it wasn’t an issue with the paver’s screed adjustment while moving but rather relying on the wrong equipment for the job – so use the motorgrader and free-hand it the last tenth or so! those crusty old timers did it for eons while walking to work uphill both ways in the rain so you can too ???
Plus GPS/DTM model builders don’t usually think like a grade foreman does or how the guy running the paver will make his pulls while moving down a street, at an intersection or through a parking lot’s short run of spaces. Some smaller contractors in parking lots are ok with the GPS/DTM model getting them close enough on the stone subgrade elevation then they’ll pull string line from lip of curb to lip of curb and have somebody check subgrade depth with a folding tape. But if you did that and the parking lot was supposed to have an inverted low for a paved swale through it that is vertically lower to say an drop inlet then would the string line method actually work from curb to curb? no probably not. But did the GPS/DTM model even reflect that intent if the engineer’s proposed contours didn’t show that inversion? my money is on Murphy’s law and the “CAD guru model builder” never having actually pulled string lines through parking lots, busted out a carpenter’s level to check fall, setup lasers for slopes or shot grade for days on end when the grades don’t work out haha – I’ll try to stop now as I don’t want to hijack a good thread on something else
Log in to reply.