Application of a hand-eye self-calibration technique in robot vision

The Lei's method is studied and analyzed from another angle in this paper. And this method is applied on the MOTOMAN-SV3X robot vision platform after being improved. Firstly, the mathematical model of the hand-eye vision positioning system is re-formulized, in which the camera intrinsic parameters matrix and the hand-eye relation rotation matrix are combined into an integral parameters matrix which is the so-called constant matrix. Then, the constant matrix can be calibrated directly by four translation motion (any three of them are not coplanar). The method doesn't require the known reference object, and it can be realized by any two image points of the environment, which reduces the error caused by solving the parameters of the two parts individually. At the same time, a new calibration method of the hand-eye position vector is proposed, where only two motions with rotation is needed, and it is simpler and higher in accuracy.