A new camera calibration method for robotic vision

摘要

When a camera is to be used to guide a robot to pick some object, one of the first thing to do is to cali- brate the camera. This consists of establishing the correspondence between the three dimensional (3D) world coordinate system and the two dimensional (2D) camera coordinate system. The correspondence is given by various camera parameters which comprise the interior orientation (focal distance, radial distorsion and aspect ratio) and the exterior orientation (rotation and translation of the camera). This problem is also known as the "interior/exterior orientation problem" or the "hand-eye calibration problem". In the present paper, the calibration problem is formulated as the problem of finding the best parameters that fit the observations of known test points. Then, a new numerical procedure is described to obtain the best parameters in the least square sense. The main result of the present paper is that the errors corresponding to each test point can be approximated by two linear equations in terms of the independent camera parameters. Then, these linear equations are combined to produce a least square matrix equation in term of the independent camera parameters. This numerical method uses only several . systems of linear equations of small size. The method presented is simple which makes it adaptable to other camera calibration problems such as stereo camera calibration for direct 3D measurements.