Hello
I'm currently playing with OpenCV to build an AR application. I'm using the findChessboardCorners() function to detect the chessboard and then estimating the pose with solvePnP(). After this I display a 3D cube with OpenGL. The rotation and translation are in the right direction.
The problem right now is, that even though the translations are in the right direciton, the distance is not correct. What I discovered is, that the size (distance between two points) of the objectPoints argument given to solvePnP() influences the results of solvePnP().
At the moment I have two things I don't understand:
- What are the units of these object points?
- How are they related to other numbers (world, image, intrinsics)?
- How can I determinate the correct size of a square of the chessboard?
I hope you can give me some hints to point me in the right direction!