UAS Topo Mapping Results

Looks good. You went way over on number of photos needed.

You could get same results, and in less time with:
Single grid Nadir flight at 200ft, as you have no buildings, no trees, and very little vertical change.
75% front and side overlap
This would be only 300 to 400 photos, and reduce your processing time to 1.5 to 3 hours.

Were the road shots on the outer edges of the point cloud, well outside of any GCP? I've noticed that the edges of the point cloud can tend to "curl" a little. I just processed about a 24 acre site and most of the ground truth points were spot on with the point cloud, but a couple points on the far south edge were out 0.2' - 0.3' in the vertical.

Did you preplan the flight lines? What software?

leegreen, post: 435742, member: 2332 wrote: You could get same results, and in less time with:
Single grid Nadir flight at 200ft,

A single grid with around 80% overlaps seems to give just as good results as a double grid. If there is any significant wind the second gris is going to end up with significantly different groundspeeds on the up and downwind legs.

As Lee D says, the edge GCPs need to be outside the area of interest to stop the "curl". Curl can happen very quickly - I've had it within 40 metres of the outer GCP.

What condition was the road surface? Where the surface is poor with small areas of the wearing course missing then those depressions can throw the matching of pixels and give rise to larger errors than you would expect.

It might be worth reprocessing without GCP13, since that appears to have greater XY variance, just to see how that compares with the original.

Yeah I know it was way overkill, but I'm still in the testing phase. I also flew a single grid at 300 feet. I haven't done a detailed comparison of those two yet.

I used Pix4D Capture for the flightplan.

Here is a map showing the location of my GCPs and checkpoints. The road is brand new asphalt in perfect condition.

Thanks!

With the DJI I'd go with a single grid, 80/80 overlap, flown at 100' - 150' AGL. I'm not sure if the camera is good enough to fly much higher than that but I guess it depends on what resolution you're looking for.

The DJI Phantom 4 Pro has a 20Mp camera with 1" sensor. More than capable at 180ft AGL for site of bare ground, as in this case of OP.

Methods = over kill. Single flight lines, @ 75%/75% would yield the same results.

Even at 20MP, the P4Pro is only good to about 250' agl for detail road work. At least in a single flight line coverage.

Thanks for the input guys.

fyi 300 ft agl results in 0.08' pixel resolution with the Phantom 4 Pro camera.

The 0.2' difference on the road shots might be due to noise in the point cloud, which is what the DSM is generated from. With photogrammetically derived point clouds and homogeneous surfaces like asphalt, I see a lot of noise in sites I have flown on roadways. You mentioned the road was brand new asphalt, so that's what I would contribute that to.

How many man hours do you think it would've taken you to compete this project with a blend of RTK and conventional equipment? How about RTK only? Just curious, mainly. I know drones are the new rage but just trying to figure out if we should even pursue the learning curve or not.

And thanks for the detailed write up!

Michael Detwiler, post: 436093, member: 12520 wrote: The 0.2' difference on the road shots might be due to noise in the point cloud, which is what the DSM is generated from. With photogrammetically derived point clouds and homogeneous surfaces like asphalt, I see a lot of noise in sites I have flown on roadways. You mentioned the road was brand new asphalt, so that's what I would contribute that to.

I assumed that was the culprit. I need to see what kind of manual editing I can do with the point cloud in Pix4D.

TXSurveyor, post: 436116, member: 6719 wrote: How many man hours do you think it would've taken you to compete this project with a blend of RTK and conventional equipment? How about RTK only? Just curious, mainly. I know drones are the new rage but just trying to figure out if we should even pursue the learning curve or not.

And thanks for the detailed write up!

That's a tough question to answer. Technically it would be impossible to generate a DSM this detailed by hand. But do you need this level of detail. The pixel resolution of the DSM is 0.05', so it has an incredible amount of detail. To think that I can click anywhere on this site and get an elevation to within 0.1' (1 sigma) is pretty amazing actually. The other amazing part is how quickly this can be produced. Less than 3 hours in the field plus another couple hours in the office produced this result. How many topo shots can you get in three hours? No way this level of detail could be replicated by RTK/conventional equipment in my opinion. Only other way would be a scanner. But you wouldn't end up with an ortho also if you used a scanner.

andrewm, post: 436124, member: 10888 wrote: That's a tough question to answer. Technically it would be impossible to generate a DSM this detailed by hand. But do you need this level of detail. The pixel resolution of the DSM is 0.05', so it has an incredible amount of detail. To think that I can click anywhere on this site and get an elevation to within 0.1' (1 sigma) is pretty amazing actually. The other amazing part is how quickly this can be produced. Less than 3 hours in the field plus another couple hours in the office produced this result. How many topo shots can you get in three hours? No way this level of detail could be replicated by RTK/conventional equipment in my opinion. Only other way would be a scanner. But you wouldn't end up with an ortho also if you used a scanner.

Thanks for the info, I wasn't trying to be negative if that's the way it came across. My business partner was certified last fall,we just haven't pulled the trigger on a purchase yet. Have you had the opportunity test it in and around vegetation?

TXSurveyor, post: 436131, member: 6719 wrote: Thanks for the info, I wasn't trying to be negative if that's the way it came across. My business partner was certified last fall,we just haven't pulled the trigger on a purchase yet. Have you had the opportunity test it in and around vegetation?

I didn't take it that way at all. I think UAS based mapping is just another tool in your toolbox.

I haven't tested it around vegetation yet, but that's my next test. But if you can't visually see bare ground, there's no magic that can happen using photogrammetry. With sparse vegetation perhaps a double grid would be beneficial to help see under the plants / trees.

andrewm, post: 436145, member: 10888 wrote: I haven't tested it around vegetation yet, but that's my next test. But if you can't visually see bare ground, there's no magic that can happen using photogrammetry. With sparse vegetation perhaps a double grid would be beneficial to help see under the plants / trees.

If the foliage is in a line rather than flying a double grid just fly additional lines parallel to the foliage and offset so that the foliage falls about a quarter way in from the image edge.
Ground vegetation is always a problem. Applying an aggressive filter to the areas should take out all the bushes, small tree groups, vehicles, etc.but the lower grasses and other plants are not so easy. I've found the best way is to run a random set of ground points through those areas, done when you put the GCP down, and use those to apply area corrections (typically around 5cm shortish grass, 1cm shin length, 30cm waist height.
Not ideal, but it is just what you do when walking a site - you look for points which appear to be typical to you and take those. The density of the DSM means that you tend to find all the small hollows you might miss on conventional survey. I normally allow about 20% extra time as the GCP will take to fix for the random walking - so I might end up with a couple of hundred vegetation points over a site of perhaps 50 ha.
If in doubt also run a string of ground points adjacent to any heavy foliage lines to enhance the aerial data.

I agree that this technique is another tool in the toolbox. My experience has been that while this technology is impressive for topographic work where it will work, I usually can do a ground survey using RTK, meeting the normal standards for an acceptable topo survey faster than I can after considering 100% of my time for each option. That is not to say that it doesn't have valid uses. When considering grass height, trees on part of the job, mobilization and demobilization with a drone, numerous control points, downtime on computer during processing, possibility of having to re-fly, etc, on most open site jobs where walking is easy, I myself cannot do a drone topo meeting normal requirements as fast as an RTK survey. That said, I find that they can be very useful in rugged terrain with little vegetative cover. However, without a lot of control points, the results can give one a false sense of accuracy with disastrous results undetected. In addition, the accuracy that Chris got does not always happen--not by a long shot--at least in my experience. Chris did a superior job setting control and checking.

Would I use a drone and say only 4 control points to shoot the volume of a quarry or stockpile or a topo and sign and stamp it for release without considerable ground checking using RTK or total station? NO, not with the present technology.

The biggest use I have obtained is the use of ortho overlay to speed up and improve the final survey drawings.

This is not to say that with the required care and time investment, drone topo won't produce a usable product on some jobs; plus it is fun and stimulates interest at work and usable products within limits.

Frank Willis, post: 436182, member: 472 wrote: Would I use a drone and say only 4 control points to shoot the volume of a quarry or stockpile or a topo and sign and stamp it for release without considerable ground checking using RTK or total station? NO, not with the present technology.

Regardless of whether you are using RTK, PPK or just GCPs anything bigger than a tennis court needs a decent number of GCP. Nine is probably the minimum. The size of site is important - unless you can't get safe access onto a site then small sites are quicker on foot. However, since 300Ha may take about 5 hours to fly and put down ground control and vegetation check points then the method is much quicker for larger sites.
As always, it is a case of assessing what can and can't be done with each tool and mixing the methods accordingly.

If there is existing marked ground control then even if the GPS signal is too poor for RTK ground survey to be successful, then as long as there are enough satellite signals to keep the aircraft on general flight track (min 6) the post processing and stitching of the images can be carried out without any image location data - just takes longer, but you have a result.

Just to add an exam,ple to my earlier post, showing when the UAV saved the day.
We had a small job, a couple of hundred yards square, which we intended to do by GPS. The customer had previously expressed some interest in aerial imagery so we decided to fly the site as well and do a comparison between air and ground survey.
On the day, for some reason the GPS was really bad, dropping out, poor signals, the lot. We just about managed to get GCP down (10 markers) and then made the decision that we would have to do the survey off the flight data only. We added in GPS checks where we could get a good response and managed to get some answer on the utility covers (to help the air survey identify where the covers were).
The end result was that the checks all tied within a few cm. with the imagery - if we hadn't had the aircraft on site we would have been on a return visit with the total station. Needed a bit of image studying and point projection to get the wall corners under the foliage!

Chris, your site was much more interesting than mine!