Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

Calibration with circle grid - center of gravity is not the circle center - is this implemented?

Hello,

my question is about the calibration methods in OpenCV. I know the user can choose between chessboard and circle patterns. For circle patterns though, there is this problem that the center of a circle is not the same anymore as the center of the circle's gravity when viewed from some arbitrary angle. As we know, objects nearer appear bigger in the camera image and objects farther away appear smaller, thus, for a circle viewed from an angle, the nearer half will seem bigger in the image than the further half and thus, the real circle center is closer to us than one might think.

This problem is discussed in this paper by Heikkila and my question now is: Is this circular center projection error considered and corrected in the calibration methods implemented in OpenCV? Or is it ignored, since the errors are somewhat small?

I'm writing my master thesis with camera calibration as a central topic and it would be very helpful for me to know the answer to this question!

Calibration with circle grid - center of gravity is not the circle center - is this implemented?

Hello,

my question is about the calibration methods in OpenCV. I know the user can choose between chessboard and circle patterns. For circle patterns though, there is this problem that the center of a circle is not the same anymore as the center of the circle's gravity when viewed from some arbitrary angle. As we know, objects nearer appear bigger in the camera image and objects farther away appear smaller, thus, for a circle viewed from an angle, the nearer half will seem bigger in the image than the further half and thus, the real circle center is closer to further away from us than one might think.think (since, for example, the nearer half takes 60%, the further half 40% of the area of the circle image, so the real circle center is at position 60%, while the center of gravity is always 50%).

This problem is discussed in this paper by Heikkila and my question now is: Is this circular center projection error considered and corrected in the calibration methods implemented in OpenCV? Or is it ignored, since the errors are somewhat small?

I'm writing my master thesis with camera calibration as a central topic and it would be very helpful for me to know the answer to this question!