Vision system calibration is the first and foremost task in ping-pong robot. To reduce the deficiency of the traditional binocular vision cameras method in heavy computation and difficulty in image synchronization

a new calibration algorithm is proposed to deal with a monocular vision system. Based on the monocular camera model，the stereo mapping is transferred to the planar mapping. Using the projection geometry relationship between light

ball and shadow

both the least square method and the optimized error estimation method are used to evaluate the parameters of the vision system. Furthermore

it lays a foundation of 3D position calculation and trace prediction. This method has been proved to have high accuracy to locate the object in the scene (<4 mm) and high-speed image processing (>60 fps). It is an effective and reliable way in computer vision compared with binocular stereo vision in structure size

measurement speed

and cost. Also it has a wide application prospect.