In the RF fingerprint identification scheme for the constellation diagram, with respect to the problem of the low identification precision rate in low signal noise ratio (SNR) environment, a two-dimensional identification algorithm based on the Euclidean distance and amplitude distance is proposed. This solution optimizes the constellation diagram to extract RF fingerprints with better identification performance from the optimized constellation diagram, and then uses the two-dimensional identification algorithm to improve the identification precision rate. Simulation results show that compared with the method based on the Euclidean distance, the identification precision rate of the proposed two-dimensional identification algorithm can be improved by up to 8% and the identification precision rate is 77.8% when the equipment capacity is 50. The RF fingerprints characterized by the optimized constellation diagram present better uniqueness and robustness.