Best calibration checkerboard config (size and square length) ?

asked 2017-12-01 13:22:39 -0500

2ros0 gravatar image

updated 2017-12-01 13:24:36 -0500


I would like to know how I should go about picking the size of a calibration checkerboard along with the length of the squares? I'm using the OpenCV calibration routine and looking to get adequate "x" "y" "size" and "skew" coverage.

I went with a default checkerboard size of 8x6 and length of square of 2.5cm but with my configuration of stereo baseline, focus distance and field of view, I don't get a good enough calibration result. I believe this is because I couldn't get enough samples in the stereo overlap region or because the "size" parameter was not satisfied.

  1. How does one pick the right calibration checkerboard size?
  2. In particular, for cameras where the stereo overlap is relatively less, what sort of checkerboard is best?
  3. Is there a way to just calibrate the intrinsics of the cameras one by one and then calibrate just the extrinsics separately?

Thank you

edit retag flag offensive close merge delete



Just on small comment on patterns: Please do not print out paper and somehow fix it to a cardboard. Invest 20$ and let it print on Aludibond: A small investment, but so much more accurate, planar and reusable than the ugly paper patterns.

FooBar gravatar imageFooBar ( 2017-12-01 13:55:58 -0500 )edit

Larger is better as it minimizes the measurement error of the calibration pattern (though it is less likely to stay rigidly planar). When the overlap is minimal (large baseline) and you don't have an enormous calibration pattern, calibrate each camera individually first and then compute the extrinsics, after having fixed the intrinsics. Also, it is my understanding that circles are better localized than corners, so you may want to go with a pattern other than checkerboard.

Der Luftmensch gravatar imageDer Luftmensch ( 2017-12-03 12:18:21 -0500 )edit