RTK/PPK checkpoint workflow

We have been using an RTK drone for the past couple of months and have found that the current gcp process does not quite work for this type of imagery. There are many occasions where we no longer want to process with ground control points but need to use checkpoints for data verification.

It would be best to have the ability to enter checkpoints and have them verified against the final surface without having to use any GCP’s.

Along with that I’m going to repeat the request to have markers available on the map automatically generated from the data that we input for GCP’s and checkpoints.


I just flew a site with an RTK drone and put in 2 checkpoints with no GCPs.
Are you saying you cant put in just check points? You can…

1 Like

I’ll have to try that again. Last time I did it that way there were no GCP’s recognized and it would not complete the process.

I actually just did this yesterday using one checkpoint and it processed it fine. Was just for testing with the new drone.

So you imported a CSV with one “checkpoint” on State Plane and orthometric height?

Yes, one checkpoint, state plane…I believe ortho height. I think that’s why the drone elevation numbers were off by the difference in geoid offset, doesn’t take that into account. You know this stuff way better than I so correct me if I’m wrong.

I’ve never tried a single-point file but am running one now and let you know. I find it hard that it would even come in within feet horizontally but we’ll see.

Well it was 84.343 feet low. Compared to my checkpoint.

That matches GEOID12B for your location which brings up a good point.

You not only have to know the horizontal datum but also the completely separate vertical datum. People still use 12B but 2018 is the latest and it doesn’t usually differ much in our area but when you are scratching for accuracy every hundredth counts. For your spot GEOID18 is 0.16ft different so add to that the fact that the offset can vary as much as a tenth across a large site you could be 3 inches off thinking that you had done everything perfectly.

This is why GCP’s have been so important and why it is a bad situation that DroneDeploy can’t seem to successfully process GCP maps with RTK-tagged images. If you are working large sites you better have at least 2 checkpoints as far as you can get them from each other. Obviously 3 would be better so you have a check.

Just tagged and submitted the checkpoint so we’ll see what happens.

So I got the results back from the single checkpoint processing and amazingly it was within 4 in horizontally which shouldn’t surprise me though being that our rovers and drones run on the same RTK network. What is not making sense is that the difference in elevation while it looks like the geoid separation is not exactly right for that location. The checkpoint was off 83.995ft but the GEOID separation at the coordinates is 85.31ft so it looks like you got some more digging to do.

One bright side of me looking at the report in more detail I found the exact proof of why DroneDeploy isn’t processing RTK imagery as well as it should be. This should be 3.28ft (1 meter) with RTK.

Hmm, the benchmark I used was shot by our survey guys elevation is 953.39 after I flew it my map elevation at the same point was 869.046, I used geoid12b put the coordinates from the benchmark in here GEOID12B - GEOID - Data XXXXXTITLEXXXXXamp; Imagery - National Geodetic Survey , there’s one for Geoid18 as well probably use that one going forward, which gave me the delta of -25.708 meters, 84.344 feet added that to my map elevation which gets me to 953.389 ft. Not sure how I figure this out just seemed logical…if I’m way off base I’m open to suggestions.
This was my first run with this new drone need to do a few more before I’m comfortable/confident with the output.
I’m not sure what average gps trust is and the help is useless but this is what I got for that flight.

1 Like

That all looks right and is how it should have worked. I am trying to figure out why my difference doesn’t match the GEOID offset for the area I ran. It looks like they got the algorithm for the M300 right. Sounds like I need to contact support and see if there is something they need to do to add the Yuneec to the more accurate image processing.