Bad Camera Calibration, then bad check points

Hi, We recently started using an Autel Evo III RTK v3 UAV for some of our mapping projects. We’re running into some strange issues. First, we had a ~39% relative difference between initial and optimized camera parameters. I thought I had that fixed by reprocessing Step 1, but when I added GCPs and check points, I’m getting pretty bad residuals on my check points. I first ran the project holding 5 of my 9 panels as 3D GCPs, and was getting residuals over 1 ft (0.30m). I added 2 panels as 3D GCPs, using the remaining 2 as check points. I’m still getting residuals of ~0.7’ (0.20m) vertical, although the horizontal is acceptable for the check points (~0.1’/0.03m). I am re-optimizing after each change in picking GCPs and/or choosing which are held as 3D GCPs and which are used as check points.

For my PPK processing of the images, I’m using a RINEX file from a Leica GS16 GNSS receiver, converted using Leica Infinity and Emlid Studio to run the PPK. I had a few float solutions, but only about 14 out of nearly 500 photos. These were at the beginning of each mission.

I noticed when I first ran Step 1, there was a downward curved “banana” shape to the calibrated camera locations. I am using the default camera settings from the Pix4DMapper database for the XL705 camera. This camera does have a rolling shutter, which isn’t ideal I know, but I don’t think that alone should have this large of an effect. We never had this issue with out Phantom 4 Pro, which is also rolling shutter.

I have attached the latest quality report.

Thanks in advance,

David

MSB_TalkeetnaDike_250522_report.pdf (1.1 MB)

Hello,

Since you have a PPK dataset, I suggest using the accurate geolocation and orientation calibration method. Mapper will place more weight on the geotags than it otherwise would. It might improve your results.