Hi,
We have fitted a survey grade DGNSS to our UAV and I’m running some tests using a basic workflow to create a DSM and an orthophoto. I’m not using control points, as I’m seeing how good we can get just with the accurate DGNSS on board.
There are cm accurate control points visible on the image (9) and some height check points (18) which I’m using for error assessment.
Given that the image coordinates are cm precise, even accounting for platform movement, I’m still getting 1m plus errors in x,y and 70cm errors in z. I’m getting less error with other SfM software.
Are there any different settings for alignment/reconstruction of precise camera positions to improve these results?
Thanks
Martin
Hi Martin,
the only thing I can come up with is to export an XMP, which includes the camera position, and then alter that manually.
Of course, that’s only possible for a small amount of images, if you can’t whip up some script for the task…
Just out of curiosity: If you have control points, why not use those for geo-referencing? It won’t take very long (I bet less than writing a script) and it’s quite helpful, also for accuracy, since with flat objects that have loose ends there is some potential for deformations at the edges.
Hi, thanks for your suggestion, I’ll have a look at XMP.
To satisfy your curiosity, I’m testing achievable accuracy without control points. I sometimes work in environments where it’s difficult to impossible to get control on the ground, e.g. salt marshes, mud flats, active land sides. That’s why we put precise GNSS on the drone.
