RTK drone images – model tilted 90° after alignment – recommended settings?

Hi everyone,

I’m working with drone images captured with RTK enabled (so the photos contain accurate georeferenced EXIF data).

When I import the images into RealityScan and run the alignment, the reconstructed model is rotated approximately 90 degrees (tilted on its side), even though the drone coordinates should be correct.

I suspect this might be related to:

  • EXIF orientation data

  • Coordinate system settings

  • Prior pose accuracy settings

  • Georeferencing or alignment configuration before processing

Before running alignment, I would like to better understand the correct workflow for RTK images.

Are there any recommended tutorials, guides, or best practices specifically for:

  • Importing RTK drone images

  • Setting prior pose accuracy

  • Choosing the correct coordinate system

  • Preventing model rotation or axis misalignment

Any advice or documentation would be greatly appreciated.

Thank you!

Hi Christina,
I suppose there are no recommended settings. Each dataset is slightly different. What were you alignment settings? Do all images have the proper coordinates? After import, are the images placed properly on the map view? Do you have also GPCs for this dataset?

All available tutorials are on our YouTube channel and there are some also there on EDC in the learning section.

Hi RS community,

First of all, thank you for the great tool, it’s really helping us at this stage(newly founded small company)!

we recently had our largest project yet, 3000+ images of local area affected by wildfire.

it was done in few flights, on different heights, with our RTK drone. Didn’t use GCP.

I initially load photos in original folders for each flight mission, cliked Align, but only third of the photos were aligned others were missing. so I put them all in one folder, Aligned them with RS again, but now duplication is visible on sparse pointcloud (lower half of screenshot).

I’m not certain if this is because banana effect or maybe incorrect alignment.

if i understood correctly by reading this forum, solution for incorrect alignment is sequencing images(align first mission, then import second, align, import third, and so on..)?

regarding banana effect, we are using Brown4 with tangential2 distortion model, I’ve read that Brown 3 or K + Brown4… can also be used. What are the differences between them, and how to know which one is the one?

i read in other topic that I can set pose of the images to Unknown prior to alignment, would that help? but that seems contra intuitive since we have RTK data from drone, it should be precise regarding its position.

i also remembered that i read somewhere that weight factors can be altered before alignment. is that possible solution, to add more(or less) weight to original positioning?

I would appreciate advice since model is large, so trial and error method in this case is rather time consuming.

Regards

Ratko