Spectral–Spatial Residual Network for Hyperspectral Image Classification: A 3-D Deep Learning Framework

TLDR

The authors propose an end‑to‑end spectral‑spatial residual network (SSRN) that directly consumes raw 3‑D hyperspectral cubes, eliminating the need for hand‑crafted feature engineering. SSRN employs consecutive spectral and spatial residual blocks that connect every other 3‑D convolutional layer via identity mappings, with batch‑normalization applied to all conv layers to stabilize training and enhance feature learning. The SSRN outperforms existing deep learning models, achieving state‑of‑the‑art classification accuracy on Indian Pines, Kennedy Space Center, and University of Pavia datasets while mitigating the accuracy decline seen in other methods.

Abstract

In this paper, we designed an end-to-end spectral-spatial residual network (SSRN) that takes raw 3-D cubes as input data without feature engineering for hyperspectral image classification. In this network, the spectral and spatial residual blocks consecutively learn discriminative features from abundant spectral signatures and spatial contexts in hyperspectral imagery (HSI). The proposed SSRN is a supervised deep learning framework that alleviates the declining-accuracy phenomenon of other deep learning models. Specifically, the residual blocks connect every other 3-D convolutional layer through identity mapping, which facilitates the backpropagation of gradients. Furthermore, we impose batch normalization on every convolutional layer to regularize the learning process and improve the classification performance of trained models. Quantitative and qualitative results demonstrate that the SSRN achieved the state-of-the-art HSI classification accuracy in agricultural, rural-urban, and urban data sets: Indian Pines, Kennedy Space Center, and University of Pavia.