An approach of calibration method for three-dimensional (3-D) vision system based on shifted point array encoding was presented in this paper. Initially a 3-D calibration benchmark was established in object space,thus the camera subsystem can be calibrated with the benchmark in light of the concept of benchmark transmission. Furthermore the projection subsystem could be calibrated according to the constraint condition between the camera and the projection subsystems. Once both subsystems have been calibrated,in light of the shifted point-array encoding method the 3-D coordinates of range image could be obtained with moderate accuracy,which could be used to compare with the prescribed benchmark in order to build up a square-error based objective function,leading to a non-linear equation. By solving this equation,all structural parameters involved in 3-D vision system could be determined with appropriate accuracies. Experiment results showed that,after proposed calibration procedure,the 3-D vision system based on point array encoding was able to achieve such accuracy as the standard deviation in X direction is 0.29 mm,0.24 mm in Y direction,and 0.29 mm in Z direction,respectively,for a measuring volume of 300×300 ×80 mm³.