Abstract: Cranio-maxillofacial region is of a complex structure, its surgery is of high-risk, and therefore it demands a high accuracy. For this reason, a set of cranio-maxillofacial surgery robot system is developed. Firstly, a spatial registration method of the cranio-maxillofacial surgery robot based on optical localization is proposed and established, to complete the spatial registration of the robot system efficiently and accurately. And the coordinate system transformation between different subsystems of the robot is realized by the spatial registration method. Then, the matrix conversion between point sets is realized by the quaternion based iterative closest points registration algorithm. The verification experiments of registration accuracy and positioning accuracy are carried out respectively, as well as the model and cadaver puncture positioning experiments of trigeminal radiofrequency thermocoagulation, based on the spatial registration of the medical image, the robot and its each subsystem. The average registration errors of registration accuracy verification experiment are less than 0.75 mm. The average error of the positioning accuracy verification experiment is 0.56 mm. The successful rate of puncture is 100% in model experiments, and it is 95% in cadaver experiments. The comprehensive experimental results show that the spatial registration method is of a high registration accuracy and a certain feasibility, which meets the needs of the cranio-maxillofacial surgery robot system.