• Hyper spectral imaging has been widely used in the remote sensing which can acquire images from hundreds of narrow contiguous bands, spanning the visible-to-infrared spectrum. • In the hyper spectral image (HSI), each pixel is a high-dimensional vector and its entries represent the spectral responses of different spectral bands. • In the existing system, sparse representation has been also applied in HSI classification, using the observation that hyper spectral pixels approximately lie in a low-dimensional subspace spanned by dictionary atoms from the same class. • In this paper the system proposed a super pixel-based discriminative sparse model (SBDSM) to effectively exploit the spatial information of the HSI.