Eon's profile - activity

I'm trying to find the simplest way to unambigously solve the camera pose (intrinsic parameters already known) from an image of a flat 2D surface with a pattern on it. I'm open on what pattern to use to make it work, here are my questions.

From my mathematical understanding there could be a couple of ambiguities when using solvePnP if all 3D reference points are planar.

First, we don't know on what side of the plane we are on. Could this be solved using an initial guess for tvec that places the camera on the correct side of the plane?

Additionally, when using a chessboard pattern that is regular, there are ambiguities due to the symmetry. This leads to additional questions. First, can findChessboardCorners handle non-regular chessboards (perpendicular rows and columns, but the squares/rectangles do not have equal area)?

Finally, is there any procedure which would help me find the correct correspondences, that is, which 2D points go with which 3D points in the non-regular case if the two sets are known? Since there doesn't seem to be any guarantee on which order the detected chessboard corners are given.