Based on the visible-near infrared spectroscopy (Vis-NIRS) technology, a new method to discriminate varieties of rice was proposed. First, the clustering of varieties of rice was analyzed by principal component analysis (PCA). Second, characteristics information of spectra were extracted by wavelet transform (WT), which as input sets for artificial neural network (ANN) to discriminate rice varieties of rice. And then a total of 180 (60 in each category) samples of three categories were adopted in this study, with 150 (50 in each category) for training sets and the remaining 30 (10 for each category) for prediction sets. The experimental results show that the identification rate reached 99.3%, which proves that the new method proposed in this study is capable to discriminate the varieties of rice with high accuracy. In addition, it might provide a new method to discriminate rice varieties.