Near infrared spectroscopy (NIRS) combined with synergy interval partial least squares (siPLS) were used to establish the rapid method for the determination of cadmium content in rice. The NIR spectra of 72 rice samples were collected and analyzed. After optimization, an approach (SG smoothing combined with MSC and mean-centering) was adopted as a pre-treatment method for raw spectrum. SiPLS method was used as extraction methods of characteristic variables, and the NIRS quantitative model was built and predicted. The RMSECV and RMSEP of the siPLS model were 0.247 and 0.261, respectively. The Rc and Rp were 0.919 and 0.895, respectively. The results showed that siPLS algorithm not only could decrease the complexity of the model, but also improved the predictive precision. NIRS can be useful as a rapid, non-destructive and convenient analytical method for primary screening and detecting cadmium-polluted rice.