This paper presents a high-precision camera calibration technique,which uses a virtual stereocalibration pattern formed by an infrared light-emitting diode(IR LED) moving along the pre-defined path on CMM.It takes into account both radial and decentering distortions and uses Gaussian distribution fitting to accurately obtain the imaging centers of feature point image coordinates.Following the maximum likelihood criterion and initial parameters are linearly computed,and final values are obtained through nonlinear minimization based on the maximum likelihood criterion.Within accurate image and space coordinates,this method can rapidly converge.Experimental results show the proposed method can satisfy the requirements for camera calibration of a vision measurement system.