47252What is the actual Field of View?
- Jan 5, 2011Hi,
I am about to set up an array of video cameras in a panorama-like configuration for a surveillance project. Given the fact that lens manufactures often states a +/-5% tolerance on the focal length, I have some concern about how one actually should mount each individual camera - precise control of the overlaps between adjacent cameras is essential. It would be best if one could measure the Field of View (hfov) for each individual camera before mounting them.
So I have used panotools (from the latest hugin release) on a sample lens and came up with the following results;
Nominal hfov = 57.5 deg (from specified focal length)
Measured hfov = 55.5 deg (from panotools)
Measured lens distortion; b = -0.010 (other variables not used)
Control points; N=665, Err_avg=0.50, Err_std=0.41 (pretty good fit)
If the measured hfov is correct, the focal length differs by 4% from nominal. [diff=tan(hfov_a/2)/tan(hfov_b/2)-1].
But is it really the actual field of view that is given by panotools in hfov? By 'actual' I mean the real-world angle measured between the leftmost and rightmost pixels in the original camera frame - i.e. what the physical camera actually sees.
I understand the lens distortion formulas described on the wiki. If the camera frame is in portrait orientation, the left-/rightmost pixels does not move during lens correction so that case should be quite trivial. But the cameras I use have landscape orientation, so one can clearly see that the left-/rightmost pixels moves during lens correction (I have used fulla.exe to verify this). My problem is that I have no knowledge of the internal chain of transformations in panotools, so I do not understand what the hfov really represents.
Is the hfov in panotools identical to the extent that the camera actually sees, or do one have to adjust it using the lens distortion formula in some way to get the 'physical' hfov?
- Next post in topic >>