Camera/GCP error disagreement

[lvargo13]

Hello,

I'm working on a project and am getting weird results using GCPs to check the error of my model that has been georeferenced using 2 different camera datasets. I have 10 GCPs that I've selected in all images in another project, and exported them as an .xml file.

1. import images
2. align images
3a. georeference point cloud with camera position data A, error is 1 m, save as one project
3b. georeference point cloud with camera position data b, error is 3 m, save as different project
* At this point, everything seems to make sense, as camera position data A should be more accurate than position data B
4a. import markers (all unchecked), while the camera position error is 1 m, the markers check error is 2 m
4b. import markers (all unchecked), camera position error is 3 m, but marker check error is 1.5 m

I'm having trouble understanding why the camera positions with the lower error have a higher 'true error' (checked with GCPs), while the camera positions with the higher error have a lower 'true error'.

Any suggestions/ideas would be great, thanks!

[Alexey Pasumansky]

Hello lvargo13,

I think you can do in the opposite way: optimize the alignment for both chunks using markers with the unchecked cameras and see the errors for the camera positions.

[lvargo13]

Hi Alexey,

Sounds good, thanks for the suggestion. But part of why I'm doing this is to estimate how well the camera positions georeference the model using the GCPs, as I have other projects with camera positions but no GCPs.

Is it correct that importing the GCPs (markers) and keeping them unchecked will give the 'true error' of the model?