 This paper proposes a new approach to remote-sensing image classification based on sparse representation of deep learning features. It uses convolutional neural networks, CNNs, to extract high-level spatial information from images, which are then represented as sparse vectors in a lower-dimensional space. This allows for more efficient classification than existing methods, such as E-NAPs and sparse coding. Experimental results show that the proposed method outperforms these other approaches. This article was authored by Heming Leon and Cheely.