"Warped" vegetation in Orthomosaic

[cadm8]

Hi SAV and sorry for the late post. Are you suggesting marking the GCPs in each photo and then running image alignment? Would you consider running a first alignment loop, marking GCPs and then re-running alignment?

[SAV]

Hi cadm8,

I would do so if I had photo alignment issues. And at the beginning I would probably not mark GCPs in every image, but only in some.

Normally I would first import the geotagged images, adjust the camera accuracy to 5m (this will depend on your UAV setup; default in PhotoScan is 10m), then convert image locations to match the coordinate system of my GCPs and then run an alignment in HIGH (Pair preselection: reference).

After that, I would import my GCPs (including x, y and z accuracies). They should already line up pretty well, hence the manual placement of markers in images will be quite easy/quick. I would normally mark each GCP in about 4 to 5 images. Then I would uncheck a few of them in the reference pane so they can be used as control points. Next I would press UPDATE in the REFERENCE PANE. Then I would manually clean outliers in the sparse point cloud and also points with a large reprojection error or reconstruction uncertainty using EDIT > GRADUAL SELECTION. Last but not least, I would run OPTIMIZE CAMERAS. Done.

A clean/high quality sparse point cloud is very important to achieve high quality datasets 'downstream' (e.g., dense point cloud, DEM, orthomosaic, etc)

If you have geotagged imagery, there is no need to run image alignment at low quality settings first (to make the manual placement of markers easier) and then again at higher settings.

All the best.

Regards,
SAV

[cadm8]

Thanks and thanks for pushing out new ideas constantly! I will apply some details you're suggesting!

[SAV]

No worries, cadm8. 
I hope the workflow works for you too.

Regards,
SAV

Thanks and thanks for pushing out new ideas constantly! I will apply some details you're suggesting!