Altitude or height (continued)

[AnnaS]

My question is a continuation from the old thread, "altitude or height": https://www.agisoft.com/forum/index.php?topic=4986.msg38769#msg38769

I used a Phantom 4 for forest surveying and am getting a large difference between my GCP and sparse point cloud elevations. Both the GCP and camera elevations are in AMSL (m). The UAV flew 116m AGL, and all camera elevations are between 3300 and 3400m. We used a Trimble RTK system with less than 1m precision to geolocate the GCPs, and these points' elevations are between 3140 and 3170m.

When I go to start manually georeferencing the images (i.e. Filter photos by marker and manually drag the marker to the visible GCP in image. Repeat 6 times or so), and then I click "Update", I see that the marker is pulled up to the sparse point cloud rather than the sparse point cloud being pulled down to the more accurate RTK coordinates. Why is this?

In the original "altitude or height" forum post, SAV responded to a similar question, advising to: "3. Uncheck all images/cameras in the reference pane so Photoscan will ignore their locations during Photo alignment and only use GCPs."

I imagine this answer could address my problem. However, I'm not sure what selections to make when aligning the photos. Here is my workflow to make my question more clear:
1) Upload photos
2) Remove images that are obvious outliers. (e.g. Estimate photo quality and get rid of ones below 0.7)
3) (I'm generating no masks, but would do that step here if I were)
4) Upload GCP marker coordinates with accuracy
5) Uncheck all images in Reference pane
6) Make sure images and markers are on same coordinate system.
7) Align photos (quality HIGH, pair preselection ??, key point limit: 40,000, tie point limit: 4,000, adaptive camare model fitting: YES)

What should my pair preselection be to create the sparse point cloud at the elevation of the markers and not of the cameras? Reference seems to be the best answer, based on the user manual. Why does un-checking the images in the reference pane affect how images align?

[Alexey Pasumansky]

Hello Anna,

Are you sure about the base level used for the GCPs and camera location measurement, that they are the same?
The difference between the camera altitude values (3300 - 3400 m) and ground control points level (3140 - 3170 m) seem to be larger than 116 meters specified as above ground level.

Also I would suggest to add markers after initial alignment is finished.


And what do you mean by saying that the markers are pulled to the sparse cloud and not vice versa? Maybe you can send the project in PSZ format with the alignment results and imported markers to support@agisoft.com so that we can check it on our side?

[Alexey Pasumansky]

Hello Anna,

The altitude information for the cameras seems to be about 90 meters above the real height above ellipsoid value, so I can suggest to correct this using the related script:
https://github.com/agisoft-llc/metashape-scripts/blob/master/src/add_altitude_to_reference.py

Also pay attention to the alignment results - in the project that you have provided there are blocks of misaligned cameras, so you may need to reset alignment for them and re-align.