A method was developed to automatically discriminate the persistent calyx fruit and fruit without calyx of fragrant pear by means of near infrared spectroscopy(NIRS). The prediction performance of differentb and regions range, different principal component numbers and different preprocessing methods of the spectra(multiplicative signal correction, standard normal variate, and derivative spectra)together with discriminant analysis(DA)was also investigated, and The calibration model was established to classify the differentkinds of fragran tpear. The research results for the fragrant pear classification showed that DA calibration models using these parameters with band regions between 9091 and 4000 cm ^-1 and original spectra are optimal, with the percentage of correct sample classifi-cation being 100% and 95% for the calibration and validation set, respectively.