Using opencv calibration function to return checkerboard finder's depth information about robot_calibration HOT 1 CLOSED

@cangjiaxaun it is certainly possible to use just the 2d pixel data of where the checkerboard is located, and this is actually what the PR2 calibration used. The issue you run into is needing many more samples to properly calibrate (especially throughout different ranges). PR2 also required some "large checkerboards" at a long distance (~3-5m from the camera) for the calibration to have any chance at converging properly.

I'm not as sure about the pose, since the pose will be determined based on the camera parameters. If you reduce the entire estimation to just the pose and camera parameters, I'm not entirely sure what the implications are (since you're certainly storing less information, and it is not clear how much of the actual error information you're losing).

from robot_calibration.