Agisoft Metashape

Agisoft Metashape => General => Topic started by: knwin on July 13, 2018, 12:31:57 PM

Title: Any Difference: processing with GCPs vs Georeferencing with GCP later?
Post by: knwin on July 13, 2018, 12:31:57 PM
Scenario (1):  Orthomosaic produced with GCP during image processing
Scenario (2):  Orthomosaic produced without GCP during image processing. Then Orthomosaic was georeferenced using GCPs in another application.

Any difference in position accuracy between above scenarios?

Thanks
Kyaw
Title: Re: Any Difference: processing with GCPs vs Georeferencing with GCP later?
Post by: JMR on July 15, 2018, 12:30:47 PM
Differences can be rather substantial. If you don't use GCP during image processing it is likely that your model will be distorted and/or badly leveled so the ortho will be wrong. It is basically a wrong approach.
Title: Re: Any Difference: processing with GCPs vs Georeferencing with GCP later?
Post by: stihl on July 16, 2018, 11:12:29 AM
JMR is absolutely correct. You will introduce relative distortions that won't necessarily show up as visual mistakes in the Orthomosaic. They'll only appear once you have some reference data to compare it with.
Title: Re: Any Difference: processing with GCPs vs Georeferencing with GCP later?
Post by: SAV on July 16, 2018, 11:33:38 AM
Hi knwin,

You should use the GCPs during 'image processing' in Photoscan because they will help to improve the so called self-calibration of your camera, which reduces the overall error of your model.

Regards,
SAV
Title: Re: Any Difference: processing with GCPs vs Georeferencing with GCP later?
Post by: cadm8 on July 30, 2018, 03:36:27 PM
Hey guys, a quick question; I recently discovered that for finished project with many photos and GCPs, I ve forgotten to deselect cameras in the reference pane. Would photoscan use the inherited better accuracy of GCPs and ignore exif data when optimizing and building cloud? Would this mistake have any effect in processing? We have had cross checks with our GPS and all surveyed points are horizontally and vertically OK

-update-: I ran a cloud compare...comparison and the clouds are identical; with and w/o image deselection...
Title: Re: Any Difference: processing with GCPs vs Georeferencing with GCP later?
Post by: JMR on July 30, 2018, 08:11:43 PM
Hey guys, a quick question; I recently discovered that for finished project with many photos and GCPs, I ve forgotten to deselect cameras in the reference pane. Would photoscan use the inherited better accuracy of GCPs and ignore exif data when optimizing and building cloud? Would this mistake have any effect in processing? We have had cross checks with our GPS and all surveyed points are horizontally and vertically OK

-update-: I ran a cloud compare...comparison and the clouds are identical; with and w/o image deselection...

Don't worry. Normally, disabling camera coordinates have very little impact in the final solution if GCP are involved. the reason is simple, GCP are weighted by default to be 2000 times more relevant than camera coordinates. So the mistake usually afects from the 4th figure after the comma. Still is a good advice to disable them because in some cases, if GCPs are badly distributed, you exif geotags may cause noticeable changes. That is particularly true if elevations are way off

Regards
GEOBIT
Title: Re: Any Difference: processing with GCPs vs Georeferencing with GCP later?
Post by: cadm8 on July 31, 2018, 02:00:26 AM
Thanks and happy to put my heart back into place!