Laser-induced breakdown spectroscopy (LIBS) combined with pattern recognition was proposed to discriminate rice species. LIBS spectra in the range of 210-480 nm wavelength from 11 different rice species were collected and preprocessed. Principal component analysis was applied to extract the characteristic variables from LIBS spectral data. Three pattern recognition methods, discriminant analysis, radial basis function neural network, and multi-layer perceptron neural network (MLP) were performed to compare the precision in identifying rice species. The results showed that the performance of the MLP model was better. The average identification rate of rice species reached 100% and 97.9% in the training and test sets, respectively, with MLP. The highest and lowest percentages for correct identification were 100% for early indica rice, Huai rice 5, Yan japonica 6, Lian japonica 8, Xuhan 1, Lvhan 1, Sheng rice 16, Yang japonica 687, and Fenghan 30, and 77.8% for Wuyu japonica rice in test sets. The overall results demonstrate that LIBS combined with MLP could be utilized to rapidly discriminate rice species.