User's spatial perception in virtual environment can be affected by many factors. Impacts of different locomotion methods on spatial perception are investigated. Differences between the subjective measurement and objective perception data on distance and orientation are recorded and analyzed from three representative locomotion methods, namely real walking, teleportation and omnidirectional treadmill walking. During the experiment, users are asked to perform tasks in three experimental scenarios, which are rationed locomotion without references, rationed locomotion with references, and navigated locomotion by references. In rationed locomotion without/with references, users move or rotate specified distances or angles by system prompts in a scene without/with references. While in navigated locomotion by references, users are navigated by ball references. The results show differences on distance perception, and no significant differences on angle perception between the three methods. Users have the best distance perception using real walking, good distance perception using omnidirectional treadmill walking, and poor distance perception using teleportation. Compared with the other two scenarios, user has the best spatial perception in navigated locomotion by references. The study provides reference for locomotion design and development in virtual reality.