In the single-axis rotation inertial navigation system with ring laser gyroscope (RLG), the single axis rotation can compensate the vertical errors of the inertial apparatus automatically, but cannot compensate the axial vertical errors, so the precision of the system is determined by the drift of the axial RLG. A novel identification method based on radial basis function network is proposed for the axial RLG drift. The inputs of the network are the latitude error and change of the temperature, and the network is trained for steady and non steady state, in which the identification capability is less than 0.0003°/h. The experiments show that this method can estimate the axial RLG drift efficaciously, and the result of the navigation is excellent and can meet the practical demand.