How much Ground Control

We always let the company flying it, determine number of targets.

I agree with Kevin, but I'd never want less than 5, and as a general rule the more the better. If you don't have them at the corners the edges tend to fall off; you need to have them at high and low elevations if there's much change, and they need to be somewhat evenly spaced.

For ground truth I like things like sidewalk corners and manhole centers where possible. If the target area is nothing but grass and/or dirt - no corners or photo identifiable objects - you may need additional targets for that.

Hold on! There has been a lot of changes in the industry since I last studied photogrammetry for my LS test. I would think that there are some aerial companies that have not kept up with technological advances in the industry, such as airborne gps and aerial triangulation who may still want a lot of labor intense physical targets and ground control. I would think knowing what other aerial companies require might be worth your while. I know there a lot of guys that do a lot of this work, so I am interested to read what they have to say. Jp

There's stuff out there; I've looked at most of it. RTK on the aircraft might help reduce the need but from what I've seen what really seems to make a difference is an IMU. It's a drastic oversimplification of the problem to just say "if you're flying x number of acres you need y number of targets". There are a ton of variable to consider, such as:

• What aircraft are you flying? What are its sensors and capabilities?
• What is the desired accuracy?
• What AGL height and with what camera and what overlap do you intend to fly to achieve that accuracy?
• Are you looking for spatial precision (only) or real world accuracy?

I'm still relatively new to the UAS game but from what I've seen and learned through experience there is no substitute for ground control and ground truth if your data needs to be accurate and fit into the real world. It just depends on what you're flying and what you're trying to accomplish.

Lee gives a very good summary. RTK and PPK can give very good results, but sometimes significant errors can creep in and the stats don't identify this. As a result the error is only found at a later (too late) stage when work is done on the ground. If you go on the ground to do site checks then you might as well use the same amount of time to put out ground targets before the flight and make sure that the PPK and RTK tie in to those. If it doesn't, then those targets can be used to force the adjustment.
I would normally suggest at least 9 points - if the site is large you need those anyway. If the site is small it isn't going to take you long to walk between them.
Target size - depends on flying height and also camera resolution. Using a Sony A6000 black/yellow boards at 40cm are fine, 30cm. are OK. White paint on tarmac - 20cm crosses show up fine. In any case I've found that a maximum spacing of 200m. is needed to keep check point errors under 5cm. with flying height of 400 ft. with the Sony. (generally much better, mean value of 1-2cm check points with a spread of around 10cm.).

Check out the ASPRS website and publication regarding geospatial accuracies. The latest update is dated November 2014, which is old given the rate of advancement in technology, but it is a good and fairly thorough standard.

Way back when in the 1970s I worked for a company that did a lot of photogrammetry work. We tried to have 9 points per frame. The project specs were usually a 60% flght-line overlap and a 15% overlap between flight lines. It was drilled into me to only select photo-identifiable points in the middle of a flat spot. That way if the plotter operator didn't get the exact spot you shot, the elevation wouldn't change enough to create issues "leveling" the old plotter.

I learned the same when I took a photogrammetry course in 1983. Four years ago I did a photogrammetric project in west central Africa using Pleiades tri-stereoscopic imagery (French satellite). I used a similar formula for ground control points. Will you be setting targets before the flight or locating photo-identifiable points after the flight, or both?

Here is an article that goes into some detail about establishing ground control points for UAS. I see that they recommend a side overlap of at least 50% rather than 15%. That may be due to the yaw, pitch and roll of the UAS being less stable than an airplane. John Hamilton wrote something a while back about survey accuracy. IIRC it was tied to the pixel resolution of the imagery.

https://dronemapper.com/uas_photogrammetry_processing

John's post.
https://surveyorconnect.com/community/threads/high-resolution-aerial-photogrammetry.329512/#post-410194

Gene Kooper, post: 437730, member: 9850 wrote: Way back when in the 1970s I worked for a company that did a lot of photogrammetry work. We tried to have 9 points per frame.

I did the same back then, but as I recall this was for only about one frame in 10 along each flight line, not every frame and only on about every fourth line. I think time has dimmed a memory or two! The flat spot rule still applies - commonsense (which appears to be lacking all too often). Nine points minimises any bulge distortion within the control area. On a UAV the area of a frame is probably about 1/100th of a fixed wing frame so you don't put 9 points on a single frame.
The side overlap of 50% is probably the absolute minimum, 65% is better. This is partly because a UAV can bounce around in gusty conditions and partly because of the turns at the end of each line. Quite often it takes a few frames before the UAV is exactly back on line, so the side overlaps at the ends can be a little suspect.

I've never done traditional photogrammetry but from what I'm reading UAS photogrammetry is a different animal altogether - we fly 80/80 overlaps.

KScott, post: 437470, member: 1455 wrote: We are providing survey data on ground control for drone flights by others. What is a rule of thumb for the number of GC targets necessary to obtain a terrain model accurate to at least 1 foot vertical? For 5 acres, 20 acres, 100 acres, 300 acres?

My initial question is why are you providing ground control for drone flights by others? That strikes me as alarming from a liability stand point, are you verifying the data you are providing? Who has the liability if there is a blunder in the processing of the data you are providing? Drones are as good as the training behind them as much as a hand gun is with a person who is properly trained to use or not use them. Sharing data by others from the liability stand point equates to giving the bullets to a rogue hand gun user who goes afoul of their training.

Who better than a surveyor to provide ground control for drone flights or aerial flyovers? Why would he have liability if they processed data wrong? He would have liability if he supplied incorrect control data, but then one should.

I have been setting ground control for aerial flights for twenty five years and just started with drone flights. Just seems obvious to me, that surveyors should be setting the control.

I don't determine where control points go, I get that from aerial company and then try to set the control points as close to their locations as possible.

Some aerial companies want a lot of control points, so aerial companies get by with a lot less, but they decide what is necessary.

We did a lidar project with an aerial mapping company in 2013, they needed only about a dozen ground control targets for 200 sq miles.

Last month we did a dump site for drone flights, maybe 110 acres total, they wanted 40 targets, so it depends, like everything else.

Kevin Olson, post: 437830, member: 9979 wrote: Who better than a surveyor to provide ground control for drone flights or aerial flyovers? Why would he have liability if they processed data wrong? He would have liability if he supplied incorrect control data, but then one should.

I have been setting ground control for aerial flights for twenty five years and just started with drone flights. Just seems obvious to me, that surveyors should be setting the control.

I don't determine where control points go, I get that from aerial company and then try to set the control points as close to their locations as possible.

Some aerial companies want a lot of control points, so aerial companies get by with a lot less, but they decide what is necessary.

I've been doing it since the 70's, the mapping company will always lay out the target patterns, then give you the allowable distances to move them, same with the drones. I would say it's only surveyors who should be doing this, in my state I would argue you need to be licensed to do it.

Thanks to all for your input.
I have no problem providing control for aerial photogrammetry and have done it many times. The mapping company has always provided the appropriate panel layout for the specified accuracy of the project.
I have never worked with a drone company before and in this case they set the GC. We were then asked us to provide control data on the points for mapping that will ultimately be used by us to design industrial facilities. I felt that the number and placement of the GC was questionable in providing the results we need and was looking for some input on if my apprehension was correct.
We have already determined to densify the control and refly a portion of the project so that our firm is comfortable with the results. At the same time the drone guys are eager to learn more about our side of the work. They are heavily invested in the technology and have a far greater understanding of the technology than I ever will and we hope to develop a mutually beneficial partnership.
Both parties have a lot to learn I think.
We will be collecting topographic data by survey methods on this initial project so that I can sleep at night, at least part of the night!

MightyMoe, post: 437831, member: 700 wrote: We did a lidar project with an aerial mapping company in 2013, they needed only about a dozen ground control targets for 200 sq miles.

Last month we did a dump site for drone flights, maybe 110 acres total, they wanted 40 targets, so it depends, like everything else.

Bear in mind that the manned aircraft can carry a high precisions GPS and IMS system, and also has $500,000 of imaging equipment, compared with the UAVs $500 camera. The distortion factor on the image processing for the UAV is many times that of a full photogrammetric survey, so a lot more control is needed to keep the residuals down.

Chris Bouffard, post: 437824, member: 12313 wrote: My initial question is why are you providing ground control for drone flights by others? That strikes me as alarming from a liability stand point, are you verifying the data you are providing?

This isn't really any different to a surveyor providing the initial control for a construction site. You put in control, you check it, you instruct the end user to carry out such checks as he considers prudent and ask him to report back promptly in the very unlikely event that something seems incorrect.
If you have a wrong GCP (or, more likely the end user wrongly identifies one of the points on the photography, which can happen if some of them are existing road markings) then it shows up in the process residuals - reprocessing without that point will confirm if it is in error. Another reason why you should always put in more than the bare minimum.

For conventional air survey control I normally put in the points as "triples" at each location, so that it obvious if one is misidentified, or if they have trouble heighting the one they are using for plan control they can use one of the others for Z.

Hi

I wrote a excel sheet to calc ground control needed based on flying height and site size.

Lost it when I cleaned my pc though. But the general rule was over and sidelap would never ever be less than 85% (some might say overkill,but UASlV cameras like the DJI are very wide so the more the better for camera parameters) and then minimum GC is 5 (0ha - 1ha) and add 1 GCP for every 1ha above that. So 5ha = 5 + 4,10ha = 5 + 9,25ha = 5 + 21,50ha = 5 + 49. After about 25-30ha we found lidar becomes cheaper than a UAV so we dont go into that territory.

Ground thruthing we do by mounting GPS to roof and doing auto survey while travelling from GCP to GCP.

Lidar side and overlap much less. We do basic 6 points and add a point for every 10ha beyond 25ha. Hope this makes sense and helps.

Just our way of doing it. Some might do it different but has given good results so far.

Someone mentioned placing GCP on flat areas. I agree it VERY important. Also what you use to avtually mark the point (talking UAV flights). Pointless using a $15000 GPS and measuring to 15mm but the point visibility on the photo is bad and the selected ppint is 50mm wrong.

We use slightly yellow premade corex boards with black dots 50mm x 50mm for flight up to 80m and same board with black crosshairs the board long 30mm thick for higher flights

Sent from my SM-N920C using Tapatalk

These people who purchase drones and Pix4D, then seek out State licensed surveyors to provide coordinate/control strikes me as somewhat similar to someone who comes into possession of a robot/schonstedt/PC/autocad/plotter. Then hires some licensed surveyor with GNSS to go around the county setting nails with coordinates for him. Next, the individual comes out the courthouse with an arm load of deeds, gets the old schonstedt/robot cranking and proceeds to locate tons of monumentation. Next, he loads up the PC/autocad and proceeds to publish pretty annotated topo maps crediting the surveyor who provided the coordinates, showing the lands he has located (mostly lands that had real estate for sale signs). Finally, this enterprising chap goes around trying to market his maps. It just seems bizarre to even write such a post, but I am truly concerned about the effect/impact of thousands of unlicensed drone mappers on the surveying/mapping profession a decade or so from now.

chris mills, post: 437932, member: 6244 wrote: Bear in mind that the manned aircraft can carry a high precisions GPS and IMS system, and also has $500,000 of imaging equipment, compared with the UAVs $500 camera. The distortion factor on the image processing for the UAV is many times that of a full photogrammetric survey, so a lot more control is needed to keep the residuals down.

Except bear in mind that many of my UAV projects involve flights at 100' or 200' rather than the thousands of feet up for a manned aircraft. That is the equalizing factor for the camera quality.

My comment was referring to the spacings Mighty Moe was comparing and explaining why the manned aircraft needed much less control. If you compare traditional manned with unmanned then the density of control may not be too far different - if measured in terms of frames between control (a single frame may cover several square kilometres).

If you are only flying at 1-200 feet then your sites are going to be small, in which case putting in some additional markers is a matter of a few minutes. They are bound to improve your accuracies and may rescue a job which has problems. Surfaces with a very even texture can quickly create errors due to the ease of mismatching pixels, as can tall grasses blowing in a fair wind, so the additional control will stop the errors being propagated over a wider area of model.