The content of amylose， crude protein and water is an important index to measure the quality of rice grain. The transmittance and reflectance spectral data of rice grains from 100 rice core germplasm resources were collected using a near-infrared hyperspectral camera， and the spectral parameters were extracted to study the method for the non-destructive testing of quality traits in rice seed. Index of rice quality components was measured using a near infrared grain analyzer after the rice kernel was shelled. A model for predicting index of rice quality was established using the spectral parameters of rice grains as independent variables and index of rice quality as dependent variables. The results showed that the modeling effect of transmission spectrum was better than that of reflection spectrum when a single spectral model was used. Combined with characteristic spectral sets of transmission and reflection， the R² of the model for predicting crude protein， amylose and water was increased from 0.74 to 0.91， from 0.40 to 0.69， and from 0.53 to 0.68， respectively. It is indicated that the modeling effect can be improved by using both spectrum of transmission and reflection， and index of rice quality can be predicted nondestructively by using the spectral parameters of rice grains.