Residuals Like Wheat in the Wind - Help Georectifying a Model from Multiple Drone Components with Differing Vertical Datums


Apologies if this is covered in an earlier post.

I am merging two overlapping Mavic 2 Pro drone RC components with data captured 3 days apart at a remote locality in Greenland. The differing GPS satellite ephemeris observed on both days has induced a 209m altitude delta between the two drone image data sets (horizontal datum looks ok). That is to say, the image EXIF files are internally consistent for altitude, but the delta between the two sets is 209m. I have an accurate set of Ground Control Points (GCPs) that I surveyed in. they are ~20m above one drone dataset and ~189m below the other.

My workflow (as a newcomer to RC - 5 days in) was to align off the drone data –> end up with two components, one each for each drone –> pull them together manually using control points (not GCPs), and then register the GCPs to align the components. I realize now this is likely the wrong sequence, and I should have started with GCPs, but I’ve invested significant effort into tightening up the model on the control points in 100’s of images. the result, pictured below has residuals that look like wheat and are pulling the GCPs towards the messed up datum. Furthermore, the model iceberged when I merged the two drone components, and I had to rotate the ground plane 45 deg to make it all ship shape and bristol fashion.

Short of manually editing the EXIF files, or starting again (nooooo…!) is there a way I can globally retain the x,y gps coords (lat/long) from the EXIF files and ignore the z(alt)? Or can I now simply ignore the EXIF files entirely with a global switch and pull everything to the GCPs without losing days of manual image registrations to the other CPs?

Component 1

Component 2

Merged Component (note clip box view due to VRAM constraints - its a 400m x 200m site (the larger box))

Merged Component with Ground Plane Re-set to Normal