okay so this has been bugging me for a while and i finally decided to post because i've read like every thread i can find and nothing is quite matching my situation.

Running a Mavic 3 Enterprise on a mining site survey, using RTK but also laying GCPs as a check. The horizontal accuracy on my orthomosaic is totally fine, well within tolerance. But the elevation model is giving me this weird doming effect in the middle of the survey area and the edges are lower than they should be. Checked my GCP coordinates three times, theyre correct. Running it through DroneDeploy right now but about to try the same dataset in Pix4D to see if the problem follows the data or the software.

Flight was at 120m AGL, 75% front overlap 70% side overlap, nadir only no oblique passes. Site is maybe 18 hectares. GCPs were all around the perimeter which i know some people say is a problem but ive done it this way before without issues. Could this be a lens calibration thing? The camera profile i used might not be perfectly dialed in, i kind of just accepted the default in DroneDeploy.

The perimeter-only GCP placement is almost certainly your culprit, especially on a site that size with nadir-only imagery. Without any GCPs in the interior the bundle adjustment doesn't have anything to constrain the vertical in the middle so you get exactly that doming artifact ? the model kind of bows upward toward the center. Ive seen this exact pattern many times. Even one or two GCPs placed roughly in the middle of the survey area usually kills it. Some people say you need a 1-per-100-acres rough rule but honestly it depends on terrain and flight params.

The lens calibration could be contributing too ? rolling shutter compensation settings if you're on the standard Mavic 3E and not the thermal version, and making sure the camera is fully initialized in whatever software handles your calibration. But id fix the GCP distribution first before going down that rabbit hole.

yeah doming from perimeter only GCPs is a known thing, happens to everyone at least once lol. also worth checking if your RTK fix was solid throughout the whole flight or if there were any float periods because mixed fix quality in the EXIF can confuse the solver. DroneDeploy doesnt always handle that gracefully in my experience, Pix4D gives you more visibility into which images got flagged.