 The paper proposes a new method for extracting effective features from hyperspectral remote sensing images, HRSI's. It uses spectral fractional differentiation, SFD, to create a feature cube which contains spatial information. This cube is then fed into various deep learning models such as SCN, 1DCNN, 3DCNN, PCA, HYBRID SN, and 3DCNN PCA. These models are compared against traditional classifiers and other deep learning models to determine their effectiveness. The results show that the SFD feature cube is more effective than traditional methods at improving the accuracy of terrain classification. Furthermore, when combined with deep learning models, the SFD feature cube can further increase the accuracy of terrain classification. This article was authored by Jinglu, Yangli, Fengxiao, and others. We are article.tv. Links in the description below.