Nadir Photographs Aligning at Two Separate Elevations / Oblique Fine

[claytonb8298]

Hello, and thank you ahead of time for the help.

We have a project that we flew approximately 60 photographs in a nadir camera orientation and 60 photographs with a 78 degree oblique camera orientation.  We used PPK to control the camera location at the time of photograph capture and all camera positions where within 6 feet elevation of each other according to the PPK results.

After importing the photographs, converting the camera positions to California State Plane Zone 5 coordinates and importing the source position PPK control file, we aligned the photographs and this is where things got weird.

All of the oblique and a portion of the nadir (about 2/3) aligned to the same elevation as the PPK source, with a portion of the nadir (about 1/3) aligned at a much lower elevation (about 50 feet lower).

More details:

For the nadir flight the first two flight lines aligned at the same elevation as the PPK coordinates and the oblique aligned camera positions.  This is the first 1 - 18 photographs.  Photograph number 19 is at the turn to the third flight line and it is about halfway in elevation between photograph 18 position and photograph 20 position.  After and including photograph 20 the remaining nadir photographs (20 - 57) are at an aligned elevation approximately 50 lower than the first 18 nadir and all the oblique camera positions.

If I process just the nadir by themselves, the weirdest thing happens, the nadir camera positions align to a coordinate position that is lower in elevation than the sparse point cloud point elevations.  As if the UAV was flying upside down inside some tunnel and was capturing photographs above the UAV or on the "ceiling" of land above it.  Which of course is not the case.

I am attaching two screen grabs, one showing the nadir and oblique aligned camera positions and one screen grab showing the nadir only camera positions.  I am also attaching a reduced resolution aerial ortho mosaic I was able to complete using the oblique images.

I appreciate any help that you can offer.  Thank you.  I have never seen this behavior before.

[Paulo]

Hello Clayton,

very strange indeed. Could you share the exported reference cameras, ticking all options to get a better grasp of your problem,
as well as PPK coordinate report?

And if you can share the data set, I could a look at it...on my side...

[claytonb8298]

Hi Paulo,
Thank you for offering to look at this.  What do you need exactly? 

I can send you the PPK coordinates, that is easy.

What do you mean by exporting the reference cameras with all ticks?

[Paulo]

Hey Clay,

I mean in reference pane, use the Export Reference Data button to export camera reference to file as follows:

[Alexey Pasumansky]

Hello claytonb8298,

If possible please also share the problematic project saved in PSZ format after the alignment of the photos to support@agisoft.com

[claytonb8298]

Thank you.  Will do.  I am working on that now.

[claytonb8298]

Here is the reference export and the PPK results.

Thank you.  Let me know what else you might need.

[Paulo]

Hello Clayton,

looking at the reference data it seems that the source coordinates were not set to PPK coordinates as difeerence between estimated coordinates and source is some 130 ft.

But these estimated coordinates area all on a similar level or height. If you want the camera coordinates to be taken into account during estimation, you need to enable them.

If you can share these 117 images, I can try and align them,,,

[claytonb8298]

Hi.  I am pretty sure we did enable them.  The first processing is without the PPK and I think that is what the export reference came from.

The delta of 130 feet is because of the difference between autonomous flight based on latitude, longitude and ellipsoid height, versus, northing, easting and orthometric height elevations (ellipsoid height minus geoid height).  But it is a pretty much constant delta between the autonomous altitude and PPK elevation.

I am not sure about sharing all the images, I am truly appreciative of your offer.  I need to wait to hear more from metashape.
CB

[Paulo]

Ok, I understand the geoid separation is around 35 m or 115 ft which explains the difference between autonomous positions (elip, height) and PPK (ortho. height)...

[Alexey Pasumansky]

Hello Clayton,

Thank you for sharing the project.

I am looking at this project now and some things seem strange to me. For example, I am checking pairs of images that have close coordinates but from different subsets (0001_0057 and 0058 - 0057-0117):
0018 and 0077 opened side by side appear to be very similar in terms of level of detail (GSD), but when you open 0019 and 0078 it looks like 0019 was taken either on lower height or with bigger focal length. And the latter effect is likely similar for all the next pairs.

If I split the calibration groups and assign 0019-0057 images to a separate calibration group and then re-align the dataset based on the same maеching points, then estimated altitude for all the cameras will be the same.
In this case the estimated focal length for the new calibration group is about sqrt(2) times bigger than initial value.

So is it possible that anything happened to the camera mode for 0019-0057 images or if they have been accidentally resized (upscaled and cropped to original dimension) during post-processing?

[claytonb8298]

Alexey,
Wow! many thanks.  I did the same investigation you did and see the same image sampling distance you noted. 

We are positive that the UAV flew at a very consistent altitude and can prove that through PPK processing.

Additionally strange, the camera used has no zoom function and the UAV was flying an autonomous preprogrammed course.

yet, what you have described is exactly what the photographs show.  We aren't sure how it happened.  Clueless and have never had this happen before.

Thank you again.

[Alexey Pasumansky]

Hello Clayton,

If you have photo originals before PPK processing, do they look the same? Also I can suggest to check the pixel size on the "upscaled" images - maybe it is a kind of digital zoom?

And maybe drone/camera manufacturer can shed some light, how it could happen. But I cannot remember anything similar with Yuneec drones.

[claytonb8298]

I am grabbing original photos from UAV SD micro card.

The PPK process does not change anything about the digital photographs.  They are not touched or changed during this process.

[Alexey Pasumansky]

Well, that is quite strange. I am not sure, how the digital zoom works for this drone, but it seems to be listed as a feature ("The digital zoom feature allows the user to digitally zoom in and out with two finger gesture") - can it be accidentally triggered on remote controller?