When implementing cooperative positioning (CP), there are a number of challenges and issues including the high computation complexity, the high probability of encountering unreliable reference nodes and coordinate uncertainty, which may offset the benefit of CP. To address these challenges, we study the impact of node reliability and propose a cooperative positioning scheme that identifies unreliable reference nodes in the system and adaptively adjusts the scoring threshold based on the geographic locations of reference nodes. When compared with other three conventional cooperative positioning schemes, our simulation results indicate that the proposed scheme can achieve the best accuracy of about 2 meters within five iterations, and it is about 32% better than the conventional schemes in terms of accuracy.