Germination rate is one of the most important quality parameters of rice seeds. In order to identify the quality of rice seeds rapidly, the hyperspectral imaging technology and the bag of visual words (BoVW) are combined to establish a grading model of rice seed germination rate. Three kinds of hybrid rice seeds, YLiangyou302, Liangyou 108 and Nei5you8015 are selected to be aged artificially for 0, 1, 2, 3, 4 d under the condition of temperature of 40 ℃ and relative humidity of 100% , and 5 dynamic gradients are obtained. Hyperspectral images of 300 samples are randomly divided into a training set (200 samples) and a test set (100 samples). After imaging selection, the germination rate test is performed and the germination rate is calculated on the 14th day. Principal component analysis (PCA) is applied to select characteristic wavelengths from the full spectral band. Scale-invariant feature transform (SIFT) is used to extract the local features of each image. All local features are clustered by K-means algorithm to generate visual dictionary. The support vector machine (SVM) classification model of rice seed germination rate is established with the radial basis function (RBF), and the discrimination accuracy reaches 95.65%. The result suggests that it is feasible to predict germination rate of rice seeds by using hyperspectral imaging technology combined with BoVM model.