The crystallinity of wood has an important effect on the physical, mechanical and chemical properties of cellulose fibers. The aims of this study were to investigate the ability of near infrared spectroscopy (NIR) to predict the crystallinity of wood cellulose and the effect of spectral pretreatment on the prediction of crystallinity in wood cellulose using near infrared spectroscopy (NIR). Near infrared diffuse reflectance spectra were collected from wood powder with a fiber-optical probe and the crystallinity of wood was determined by X-ray diffractometer (XRD) in this experiment. The results showed that near infrared spectroscopy coupled with partial least square (PLS) regression could be correlated with the crystallinity of plantation wood, and the ability of NIR prediction based on original spectra was better than that based on the first derivative or second derivative treated spectra. There was a significant correlation between NIR spectra and XRD determined crystallinity with a correlationcoefficient of 0.950 and a low RMSEP. Near infrared spectroscopy coupled with multivariate data anlaysis has proven to be an accurate and fast method for rapid prediction of wood crystallinity.