Revision history [back]

For calibrating the intrinsic camera parameters of a thermal camera I had satisfactory (if not perfect) results by warming the checkerboard pattern using a high powered lamp. The black squares will get hotter than the white squares and show up differently on thermals after a short while. I did that with the Matlab toolbox where one has to manually click the correspondences though, not sure how OpenCV calibration copes with sometimes not so perfect images (due to not perfectly homogenous heating, low resolution, noise in the framegrabbed images etc). We also considered making a calibration pattern out of metal squares, but ditched that idea because the "paper heating" method worked well enough.