Self-Calibration Technique Based on Hand-Eye Vision Systems

A flexible valid self-calibration technique for robotics hand-eye systems is proposed. It is not necessary to disjoin the camera intrinsic parameter matrix and the orientation matrix, so they can be obtained as a "black-box", where every parameter of every part needn't be computed, thus the computation cost and the error are both depressed. In the process of calibration, the known reference objection isn't needed, and it only requires the camera to observe images from couple arbitrary reference points at four (or more) pure translational motions and two (or more) random motions with revolving, respectively, thus the proposed self-calibration of hand-eye systems is implemented, where the motion times is decided according to the given criterion. Furthermore, there is no restriction to the installation and the motion direction of the camera, so the proposed technique is easy to use and flexible.