Agisoft Metashape

Agisoft Metashape => General => Topic started by: smescarzaga on September 17, 2018, 06:30:21 AM

Title: Wild differences in elevations between nadir only and nadir/oblique mix
Post by: smescarzaga on September 17, 2018, 06:30:21 AM
The title pretty much sums it up. Let me give some background info on the data, processing etc:

The imagery from a high-end survey camera and delivered along with post-processed camera coordinates and omega/phi/kappa values. Manned-aircraft, flying at ~ 2333m (height above ellipsoide, I think since it was delivered in WGS) collected 0.25m/pixel nadir and obliques (forward and aft) at each camera coordinate. Although photos within the oblique collection are smaller (width x height in pixels and also dpi), I'm using the same camera calibration values that I used for the nadir data.

I was keen to include the oblique dataset in the processing because there wasn't great overlap with just nadir alone (I don't believe 3D reconstruction was the primary purpose of the survey). Comparing results from nadir only and nadir/oblique processing runs are...interesting. I'm getting better overall error with the nadir only job, which I guess I understand. HOWEVER, when looking at elevations values between the two, they're waaaaay off. Nadir only values are up to ~200m off from expected values whereas the nadir/oblique mix produces much more reasonable values.

Can anyone tell me what gives and how to fix it? I've read in the manual that PS expects camera heights to be height above ellipsoid, so I'm hoping the values that were delivered with the photos are ellipsoid height as opposed to have some geoid applied to it. I can't see this mix up creating such a large discrepancy though.

Thanks!
Title: Re: Wild differences in elevations between nadir only and nadir/oblique mix
Post by: JMR on September 17, 2018, 10:17:04 AM
what setting did you use in the ref panel?
How did you convert and fixed camera calibration parameters?
Title: Re: Wild differences in elevations between nadir only and nadir/oblique mix
Post by: smescarzaga on September 17, 2018, 09:44:17 PM
Attached is a screenshot of the reference settings panel. These are used in both nadir only and nadir/oblique mixed.
The photos came with no exif data, only a recent camera calibration document from the manufacturer (from which I got focal length and pixel size).
Title: Re: Wild differences in elevations between nadir only and nadir/oblique mix
Post by: smescarzaga on September 17, 2018, 09:50:00 PM
I should also attached screen shot of camera calibration here.
Title: Re: Wild differences in elevations between nadir only and nadir/oblique mix
Post by: Alexey Pasumansky on September 18, 2018, 07:33:45 PM
Hello smescarzaga,

Maybe you can also share the PDF reports from this project processed using nadir-only and nadir+oblique images?
Also do you have the camera calibration certificate?
Title: Re: Wild differences in elevations between nadir only and nadir/oblique mix
Post by: smescarzaga on September 18, 2018, 11:59:02 PM
Hey Alexey,
Thanks for responding. I'll message you the reports from both nadir only and nadir/oblique jobs (they are chunked up differently only due to more images being processed in the nadir/oblique job). I'll also message you the calibration certificate.
S
Title: nm
Post by: WSurvey on September 19, 2018, 08:43:10 AM
nm
Title: Re: Wild differences in elevations between nadir only and nadir/oblique mix
Post by: smescarzaga on September 21, 2018, 10:58:31 PM
Hi folks,

I just wanted to check in to see if anyone had any ideas as to why this is happening. It seems strange that cameras GNSS/IMU hihg-precision data is producing such strange elevation values.

I'm also attaching here one of the .geom files that accompanies each .jpg. I've stripped out lat, long and altitude from the line
Code: [Select]
latlonh_platform_position omega, phi, kappa from their respective lines. This .geom file is from a nadir shot.

Is there anything in here I could potential add to the processing to help constrain those z values?
Title: Re: Wild differences in elevations between nadir only and nadir/oblique mix
Post by: smescarzaga on September 28, 2018, 01:29:55 AM
For anyone still interested: After speaking to a photogrammetry expert, it was mentioned I should fix the initial focal length before aligning and avoid optimizing it during gradual selection. This seems like it's actually producing a more accurate ground elevation.

I'd still like to input the principle point coordinates and k1, k2 and k3 values listed in the cameras calibration certificate. The only problem is I don't understand which units PS wants these values input as.

Attached is a screenshot of the calibrated parameters of the camera and of the radial distortion factors. Can anyone tell me how to input principle point coordinates and k1, k2 and k3 values into PS properly?