Has anyone else experimented with the fewest number of circles that OpenCV will detect on an asymmetric circles grid? I'm working with a low resolution camera, so I need a large, simple fiducial for long range detection.

I printed a few quick different tests, and the minimum detection I could get was a 5 rows by 3 across grid. For reference, this example target is an 11x4.

Ideally I could detect something much smaller such as a 3 x 2 or even a 3 x 1 (a 45-45-90 triangle).