A method of classification based on hierarchical sparse representation feature learning as hierarchical discriminative feature learning algorithm is developed for hyperspectral image classification. The spatial-pyramid-matching model is used, and the sparse codes learned from the discriminative features are obtained by max pooling in each layer of the two-layer hierarchical structure. The representation of features achieved by the proposed method are more robust and discriminative for the classification. The proposed method is evaluated on two hyperspectral datasets, and the results show that the proposed method has good classification accuracy.