What's the minimum depth a stereo camera can "see"?

Question

I'm interested to find the closest depth a (perfectly aligned) stereo camera can measure (with a fixed baseline and focal length). I know that the depth can be calculated given the baseline, focal length and the disparity. Depth = (Baseline * Focal lenght) / Disparity

This problem can (in my opinion) also be formulated like this: Find the maximum disparity possible of the stereo vision camera. As Z would go to zero, the disparity becomes infinity. But this can't be true because (again, in my opinion) it's not possible to picture an image at depth 0.

Any suggestions?

Thanks in advance!

Answer 1

In a practical stereo setup the minimal resolvable depth is determined by either of two factors:

• The optics, specifically the nearest distance at which the left and right images are well focused.
• The overlap of the cameras' fields of view. As cameras have finite size, the baseline can shrink only so much, so that even with cameras toed-id (i.e. with converging focal axes) there will be a minimal distance below which some part of the scene of interest object is visible in neither camera, or in only one of them.

These constraints apply to any stereo rig. The absolute limit of a particular rig depend, of course, on its own design: on one hand of the spectrum, you can do stereo reconstruction with a pair of microscopes; on the other hand, stereo telemeters for naval artillery in WW1 could already reliably measure distances to the meter resolution at 20 Km distance.