Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

OpenCV calibration vs. Matlab calibration accuracy

Hello, So I've programmed a simulation to project a plane onto a flat screen. image description image description

The red circle is my "camera", the blue is the object being projected, and the green are the 2D image point projections. So, I know exactly where my camera is in my simulation. I used OpenCV calibrateCamera() and used the returned camera matrix to find the camera location coordinates, and then compared that with the actual one, and my normalized distance error is .000301162. I tried increasing the amount of iterations for my term criteria, but it just converges at this number. Then I also used MATLAB estimateCameraParameters(), and my normalized distance error was 1.76E-12.

To summarize: Source error for OpenCV = .000301162. Source error for MATLAB = 1.76E-12. This is a significant difference, and I was wondering if there were any settings or additional functions in OpenCV that will make my values turn out the same as MATLAB's, and if there is any reason to why this is like this.