On the possibility to add more camera calibration models

[Benny]

Hi,

When using imagery coming from DJI drones, we always face the issue of a camera calibration of limited quality. This is highlighted by the residual plots, which show remaining radial residues that are not explained by the Brown's equation.

Recently, a team of the Technological University Dresden proposed a bi-radial model for lens distortion calibration. The paper is here: https://www.mdpi.com/2072-4292/15/22/5283. This model would work well for images coming from the Phantom 4 series or the Mavic 1 or 2 series. It may work for the Mavic 3 Enterprise series, but the residual plots are much more complex, with photogrammetric results containing significant (i.e., non negligible) errors on the sides of the reconstructed areas. By the way, this is really surprising for this new series of drones dedicated to professional users.

Having the possibility of using more complex distortion models would be a real added value for Metashape Pro compared to other commercial software. The possibility to add our own camera model would even be a greater improvement.

Thank you in advance for considering this potential improvement for Metashape Pro.

Best regards,

Benoît Smets

[ilia]

Rumors are that some professional grade Mavics apply some correction to images and this is why resulting model may look like that.
Suggestions are:
a) Use raw images
b) Process raw images without any profiles in Adobe Camera Raw or any other raw tools.

Would be nice to hear if anyone can confirm that. Applying incorrect distortion model may give such weird behavior at the end when the resulted image don't follow well the basic distortion model because incorrect one was already applied.