Settings to process RTK data accurately.

I have a question about RTK and getting accurate results in processing. I use a Mavic 3E and have an NTRIP subscription that gives corrections in NAD83/NAVD88 and I process in Pix4D. I also use an Emlid RS2 and have surveyed my own ground control. My corrections come in as NAD83. I know my Emlid RTK data is accurate and settings are correct. I have had an issue getting my z axis to line up correctly with my control, typically being about 3 feet off after processing in Pix. I’ve tried reprocessing a test map with multiple different settings, from geoid offset to the different types of NAD83 that Pix has to offer and still haven’t been able to lock in my settings to make RTK reliably accurate and simple. Anyone know of a way that I can natively process through Pix with consistent settings, and have an accurate RTK map? Thanks!

1 Like

I have had a similar experience with the 3E. Horizontal accuracy is good, but vertical is off. There doesn’t appear to be any way to get good vertical accuracy without using ground control. I don’t think the problem is linked to Pix4D either. Both Propeller and Drone Deploy will require some sort of elevation adjustment to produce an accurate map. There seems to be an issue with the ellipsoid heights that the drone is outputting, or the IMU data, or something like that.

I’ve seen the elevations be off from .5 feet to almost 4 feet. It does not make a difference if it’s PPK or RTK. It’s different from flight to flight and elevation. The relative accuracy seems to be good which is why some programs can get by making the same elevation adjustment across the entire dataset. The solution seems to be using 3 GCPs in Pix4D to handle the correction. 3 GCPs because it’s the minimum for Pix4D. Not ideal, but far less GCPs that you would normally have to set out.

One item I can suggest is to make sure you are using the latest version of PIX4Dmapper. After version 4.8.3, the camera database was adjusted for the Mavic 3 Enterprise. Prior to this, the vertical elevation for the images would default to EGM96 when it should have been ellipsoidal.

1 Like