The aim of this research was to establish the fast detection method for distinguishing the rice samples from different geographical origins and different breeds in China by using Raman micro-spectroscopy. A total of 123 rice samples from Heilongjiang, Jiangsu and Hunan province were analyzed by Raman spectra, and the data were statistic analyzed by chemometrics. The data was subjected to principal component analysis (PCA) and partial least squares discriminant analysis (PLSDA) to distinguish differences among samples from different geographical origins and different species. PCA could classify the rice samples preliminarily and classification model developed by PLSDA was used to classify and predict different rice samples. The rice samples (2/3) were as a training set of modeling, and the rest of samples were as a prediction set of modeling. The correct classification rates in the training set according to the varieties of rice samples were 100%, and those prediction set were 100%, 100% and 94.12%, respectively. The results in this research indicated it is a quickly efficacious method to identify rice from different geographical origins and species by Raman spectroscopy with chemometrics.