Geo-tagging Accuracy

I am busy working on my masters thesis which uses drones to track something down below using time lapse images. I.e. take the photo into AutoCAD and scale based on its height.

What i’m finding is that the Geo-tagged photos seem to drift around quite a bit, ±0.5m. I was wondering who else has a similar application and if there is any papers/case studies on the drone geotagging rather than taking the entire set and generating a photogrametric model (which i assume smooths out the error over all of the photos in the set to get a much better accuracy than the drone GPS).


Welcome back @Ryan_RyRy. First off congrats on doing a thesis involving drones! Can you expand a little on ,

What AutoCAD? Plain or one of the other Autodesk products?

Due to the limitations of the native drone GPS module the spec’d accuracy is 1-3m so you are doing good. Only a drone with a controller that can function on an RTK GPS system.

Thanks @MichaelL

I am looking at tracking a floating orange. So hover 30m above and see the movement between frames. Hence its in water and there arent really any ground control points. The other variation is the yaw direction.

I have Autodesk AutoCAD 2019 Student Version which i have built a macro for to import the photos.

Do you have a reference for the drone GPS spec’d accuracy. Ive been testing with a Mavic Air.

I just found the new P4RTK which looks like its more suited for the job. Are you awair of any altertatives.

As an asside i wonder if a phone GPS could provide a similar RTK function especially as i would like an error less than 10cm and the 1cm of the P4RTK may be overkill.

I have also noticed that the error is much worse when the drone moves, if it is stationary then its not so bad.

I am not clear on what this is, but from what I can gather maybe you can make a shaped buoy? Something like a pentagon that will allow you to track multiple points.

The specs on the mavic air are 0.5m vertically and 1.5m horizontally. when you put on a setup like the emlid than the GPS of the drone is disregarded and you get the accuracy to geotag the photos with the Base and Rover data.

For the GPS on the drone I am having good results from the Emlid Reach setup that I have been testing. In my opinion there are some things about the P4RTK that don’t add up to the solution that they sell it as. Number one being that it is probably not a good idea to rely on rtk when flying. A few small outages will destroy a map. I prefer to use PPK. This allows you to post-process the data collection from the drone and a base station independently and then sync them together for an RTK solution.