calibration

edit

I think I completely lost now. I don't know why we need to do calibration. After we done calibration, we should be able to convert any pixel in images into millimetre in the world, that's why we do calibration, right? but using the opencv toolbox, we know the camera intrinsic parameters, and different calibrated chessboard position (rotation and translation),but if we put a new chessboard in, and get the image of it, and know the corners of it, but we still can't know , say, each corner points on the new chessboard in the real world, then what is the point to do the calibratioin? As my understanding, after the calibration done, we should know the projection matrix from world to image, and convert points in image to world.