In order to increase absolute positioning accuracy, a 3D-vision sensor which can be integrated into machining robot is introduced and a calibration method to determine the position and pose relationship between 3D-vision sensor and robot is proposed. It only requires the sensor to scan a standard sphere with known radius at several different robot poses and then the transformation matrix between 3D-vision sensor and robot can be optimized by a two-step algorithm. First, the solution of transformation matrix can become a constrained least square estimation problem, which is solved by Lagrange multipliers. Then, the transformation matrix together with pose data of robot involved for solving can be further refined by a nonlinear optimization procedure. The experiments are conducted on a 5-DOF hybrid robot named TriMule to test the validity of the proposed method.