Shallow-to-Deep Spatial–Spectral Feature Enhancement for Hyperspectral Image Classification

Since Hyperspectral Images (HSIs) contain plenty of ground object information, they are widely used in fine-grain classification objects. However, some objects similar and the number spectral bands is far higher than categories. Therefore, it hard to deeply explore spatial–spectral joint features with greater discrimination. To mine HSIs, a Shallow-to-Deep Feature Enhancement (SDFE) model three modules based on Convolutional Neural Networks (CNNs) Vision-Transformer (ViT) proposed. Firstly, containing important information selected using Principal Component Analysis (PCA). Secondly, two-layer 3D-CNN-based Shallow Spatial–Spectral Extraction (SSSFE) module constructed preserve spatial correlations across spaces at same time. Thirdly, enhance nonlinear representation ability network avoid loss channel attention residual 2D-CNN designed capture deeper complementary information. Finally, ViT-based extract (SSFs) robustness. Experiments carried out Indian Pines (IP), Pavia University (PU) Salinas (SA) datasets. The experimental results show that better can be achieved by proposed feature enhancement method as compared other methods.