Multi-viewpoint optical positioning system is an effective solution to make up for the disadvantages of light occlusion in optical positioning because it can reduce the blind area in the operation and obtain a larger field of vision. In this paper, multi-viewpoints optical positioning algorithm based on the optimal reconstruction accuracy is proposed to solve the problem of light occlusion in optical positioning. First, the proposed algorithm analyzes the number of optical positioning markers in each viewpoint to confirm the light occlusion of each viewpoint and surgical instrument. Then, according to parallel stereo vision measurement accuracy analysis model, the influence of the multi-viewpoints structure and the position relation between the marker and the camera on the measurement accuracy is analyzed. Finally, the viewpoint pair with the best reconstruction accuracy and no occlusion is selected surgical instruments. Experimental results indicate that the proposed algorithm can achieve accurate positioning and real-time tracking of surgical instruments when there is a certain view occlusion.