Abstract:In recent years, deep learning-based methods have exhibited remarkable performance in the field of hyperspectral image (HSI) classification. However, conventional supervised methods heavily rely on a substantial number of labeled samples. Self-supervised learning, as a prominent unsupervised representation learning technique, offers the potential to extract valuable information from unlabeled data. In this article, we introduce a novel unsupervised approach called self-supervised learning with the multiscale densely connected network (SS-MSDCNet) to make full use of unlabeled samples for HSI classification. First, a two-stream structure was designed to generate more positive pairs, which enables the contrast self-supervised training to learn more useful information from unlabeled data. Subsequently, a data augmentation technique based on spectral splitting was proposed to coordinate the two-stream structure of SS-MSDCNet, enhancing spectral information expression. The backbone of the proposed approach is the multiscale densely connected network (MSDCNet), which obtains input HSIs with various spatial scales by removing peripheral pixels from the original input and subsequently extracts multiscale spatial-spectral features using 3-D densely connected modules and 3-D spatial attention modules. The 3-D densely connected module effectively harnesses multiscale features extracted by various convolutional layers, while the 3-D spatial attention module enhances the network's focus on features conducive to accurate classification. To validate the efficacy of our approach, we conducted extensive experiments using four distinct HSI datasets. The results unequivocally demonstrate that SS-MSDCNet outperforms several well-established supervised and unsupervised classification methods. Furthermore, we designed a transfer experiment to confirm SS-MSDCNet's robust generalization capabilities. The code is available at https://github.com/mrblank99/SS-MSDCNet.

Multiscale Random Convolution Broad Learning System for Hyperspectral Image Classification

Random Convolutional Network for Hyperspectral Image Classification.

RMCNet: Random Multiscale Convolutional Network for Hyperspectral Image Classification

Multi-Stage Convolutional Broad Learning with Block Diagonal Constraint for Hyperspectral Image Classification.

Multiscale Spatial-Spectral Convolutional Network with Image-Based Framework for Hyperspectral Imagery Classification.

A Multi-Kernel Mode Using a Local Binary Pattern and Random Patch Convolution for Hyperspectral Image Classification

Hyperspectral Image Classification Based on Multiscale Cross-Branch Response and Second-Order Channel Attention

SPECTRAL-SPATIAL MULTISCALE RESIDUAL NETWORK FOR HYPERSPECTRAL IMAGE CLASSIFICATION

Broad Graph Convolutional Neural Network and Its Application in Hyperspectral Image Classification

Hierarchical Multi-Scale Convolutional Neural Networks For Hyperspectral Image Classification

A Novel Classification Framework for Hyperspectral Image Classification Based on Multiscale Spectral-Spatial Convolutional Network.

Multiscale Densely-Connected Fusion Networks for Hyperspectral Images Classification

Hyperspectral Image Classification Based on A Multi-Scale Weighted Kernel Network

Multiscale DenseNet Meets With Bi-RNN for Hyperspectral Image Classification

Self-Supervised Learning With Multiscale Densely Connected Network for Hyperspectral Image Classification

Bidirectional-Convolutional LSTM Based Spectral-Spatial Feature Learning for Hyperspectral Image Classification

Hyperspectral Image Classification Based on Multiscale Spatial and Spectral Feature Network.

Multi-scale 3D-convolutional neural network for hyperspectral image classification

Hyperspectral Image Classification Based on Spectral Multiscale Convolutional Neural Network

A Novel Classification Framework for Hyperspectral Image Classification Based on Multi-Scale Dense Network

Hierarchical Broad Learning System for Hyperspectral Image Classification