Vision: Internal and external camera parameters

[joptimus]

Hello,

I please need your help to assess if what I plan is possible using LabVIEW Vision:

With a camera I'm looking at a calibration grid of evenly spaced dots. There are two kind of errors that can occur in that setup:

1. Distortions from the camera optics -> internal errors
2. Distortions from perspective errors (not looking under a perpendicular angle) -> external errors

There are two problems:

1. As far as I know, LabVIEW camera calibration either accounts for perspective-only errors or perspective AND distortion errors. There seems to be no function to only correct distortion errors of the lens? See also this thread which provided only an external MatLab solution:
   https://forums.ni.com/t5/LabVIEW/Camera-calibration-internal-parameters/m-p/719433#M329220
2. I don't want to correct perspective errors, but to determine the positions of the camera (5 degrees of freedom: all 3 euler angles and translation in a parallel plane relative to the calibration target. The distance to the target remains fixed) by taking pictures of the (fixed) calibration grid under different viewing angles. Is that possible with Vision? Especially to clearly differentiate between translational and rotational errors (and combinations thereof).

If it helps, a short description of the underlying use case/application:

We're looking at a virtual display from different angles. There are always some distortions in the image which vary depending on the viewing angle. These distortions have to be analyzed, not corrected. It is important, that the camera positions for those specified viewing angles is as exact as possible.

The idea is to find out the position of the camera with a fixed and perfectly positioned calibration target.

[Bob_Schor]

Wow!  Both tasks (a comprehensive LabVIEW utility to handle camera/lens calibration and compensation, and a routine that analyzes a known image and determines camera position/orientation) are very worthy, and potentially challenging, tasks.  I haven't closely examined the Caltech MatLab code, but it looks like (a) the problem has been "solved", in principle, but (b) there is a considerable amount of code that must be developed and tested.  As for the Position/Orientation question, I can imagine the necessary math (and could probably do it), but I've not played with this level of Image Analysis.

This might make a very appropriate Masters-level Engineering Project.  I volunteer to be an Outside Examiner ...

Bob Schor