depressed /sunken surface not recognized during camera alignment, why??

[kiki]

Hi,
I am trying to generate a pointcloud of a restraining fault bend on a clay surface which will have a depressed/sunken area right in the middle. However, when I add the markers and align the cameras. Metashape does not pick up points in the depressed section of the clay surface (it is to say, I only get gray dots instead of blue ones which are the ones used for the point cloud generation in that area). I have tried playing with the key points and tie point limit on the align pictures window to try to get as many points as possible during the alignment but it doesn't work for the sunken area. Does anybody knows why this happen? Does metashape not like negative Z values? or how can I fix this?

Attached is the onedrive link to the folder that has the pictures (only 5 pics) of the surface that I am trying to recreate, CSV for the markers, and the metashape file that I worked on.
 https://umass-my.sharepoint.com/:f:/g/personal/cramossanche_umass_edu/EqAiRclSlH5FmdsQG2iDQ3cBHKcSHmhChcpQ_DHCgTxMrw?e=Cp7b9a


thank you!

[Alexey Pasumansky]

Hello kiki,

I see that you are using fixed pre-calibrated parameters for the cameras - how they were obtained?

Also please specify, if there was a multi-camera setup, or whether the photos have been taken by the hand-held camera? I would say that the focal depth is not sufficient for some photos and the tie points are not detected in large areas for such images, also ISO used is unreasonably high and affects the image quality.

[kiki]

Hi Alexey,
Thank you for your reply.
It is a multiple camera set up in which we use 5 different cameras (on tripods) for all the angles shown on the pictures. We calibrated all cameras with their zoom lenses by following the instructions on the manual and using the chessboard pattern given on metashape right before we ran the experiment. So the calibrated values you see are the ones we got after calibrating each camera with the chessboard.
Also, what do you mean by ISO? Is it the key point and tie point parameters used?

Thank you!

[Alexey Pasumansky]

Hello kiki,

I have tried to reprocess the data: https://www.dropbox.com/s/eejhp975zdx6ov2/T6-reprocessed.psz?dl=0
in the first chunk the same tie points as in your project have been used, and in the second chunk - newly detected tie points (reprocessed from scratch). However, I have disabled pre-calibration results, as some values do not seem reasonable to me. Not sure, how accurate could be calibration of zoom-lens, when focal length is set to intermediate value and the lens it not taped to physically fix the focal length.

If you are planning to continue using the same approach for new projects, you may consider adjusting the focal length of the camera lens to min or max values (since you are using zoom-lens), as it could help to avoid accidental changes of the focal length due to some unexpected movement.

Since the cameras are static, you can set F/stop value to F/8 - F/11 and longer shutter delay, while ISO setting for the cameras can be minimized at least to 100-200. It should help to get better images.