The best way to do it would probably be to use a fixed lens and use C spacers to make the focal length short enough to image a calibration pattern. A 3" pattern is about $600 and is accurate to less than a micron. Then once the lens is calibrated remove the spacers so you can image the curve. You can't use a zoom lens because changing the zoom changes the error mapping of the lens.

Like you said a high altitude is better because more curve increases your measurement to error ratio. Anyway there's only a handful of people who know how to do all that but someone could probably research it and figure it out using something like OpenCV.

Also, after all that some model for atmosphere refraction should also be considered.

So in the end you cannot just look with your eyes and think something is flat. Imagine printing out 10 lines on paper and one has a 1% curve. You could never see the difference.

The best way to do it would probably be to use a fixed lens and use C spacers to make the focal length short enough to image a calibration pattern. A 3" pattern is about $600 and is accurate to less than a micron. Then once the lens is calibrated remove the spacers so you can image the curve. You can't use a zoom lens because changing the zoom changes the error mapping of the lens.

Like you said a high altitude is better because more curve increases your measurement to error ratio. Anyway there's only a handful of people who know how to do all that but someone could probably research it and figure it out using something like OpenCV.