Hyperspectral Image Classification Based on Fusing S3-PCA, 2D-SSA and Random Patch Network

Recently, the rapid development of deep learning has greatly improved performance image classification. However, a central problem in hyperspectral (HSI) classification is spectral uncertainty, where features alone cannot accurately and robustly identify pixel point image. This paper presents novel HSI network called MS-RPNet, i.e., multiscale superpixelwise RPNet, which combines superpixel-based S3-PCA with two-dimensional singular spectrum analysis (2D-SSA) based on Random Patches Network (RPNet). The proposed frame can not only take advantage data-driven method, but also apply to efficiently consider more global local knowledge at super-pixel level. Meanwhile, 2D-SSA used for noise removal spatial feature extraction. Then, final are obtained by random patch convolution other steps according cascade structure RPNet. layered extraction superimposes different sparial information into multi-scale features, complements various land covers. Finally, fusion classified SVM obtain results. experimental results several datasets demonstrate effectiveness efficiency outperforms current state-of-the-art methods.