Distance errors for constraints in large terainn reconstruction

[xsherlockpl2]

Hi,
I’m reconstruction about 25hectares of land, with 1800 photos captured by DJI Air2.
I made those during 2 days, with both datasets of about 900 photos each as I was running out 3 batteries. A combination of topdown photos and grid 45deg oblique. Each dataset on its own aligns perfectly with distance errors < 3cm for the ground markers I placed.

Unfortunately when I try to align whole data set the error for the distance constraints go very high. And the overlap between the 2 day flights is quite high, at least 200 pictures cover the same area. and I have 2 control pints there.

And there is some strange skew in the resutls with TB-TC distance having 39cm vertical diffrence.

And the court is perfectly flat. So that is wrong, I I only do 2nd day that has most of the pictures of the court the result is perfectly level.

Is there a way to enforce the that TA,TB adn TC are on the same altitude? I do not have RTK gps to measure the proper GCPs there so I rely on what RC witll figure out and for most cases it just fine, Just for those 2 data sets I somehow introduce this error that I cant get rid of

here are my aligment Settings

I would appreciate any ideas how to fix that.

And here are my pics that you can see are quite off after alligment

[kumateCR]

Hello,

Were the images taken under the weather conditions? Did you modify images before using them in RealityCapture?

After importing all images, group them. To do this, click on the Images in the 1Ds view and use the “Group” button in the Inputs panel:

There are also some tips for large drone sets here:

Banana Effect - What To Do If My Model is Bent | Tutorial

[xsherlockpl2]

The images were taken on 2 days, under similar lighting conditions.

Every day aligns on its own perfectly with no distorsion.

800 on 1 day and 800 photos on the other day

if I add pics from the second day to the aligned project the error for all distance constraints will go from 3cm to 30cm.

I admit that I have not tried to add the second day in batches, of like 200 or so pictures , would that make a diffrence?

[kumateCR]

When taking images with drones on separate days, the issue could also be in the geolocation information. it could have been that the GNSS did not initialize correctly or fully, so the position accuracy may differ.

Disable absolute pose for the images from the first OR the second day, and leave only one-day images for georeferencing.
To disable the absolute pose, select all images (I recommend importing day 2 images and pressing CTRL + A, changing the absolue pose and then importing day 1 images) and in the Selected inputs panel under the prior pose change the Absolute pose to Unknown.