Multi-Scale Dense Graph Attention Network for Hyperspectral Classification

In recent years, numerous deep learning-based methods have gained increasing attention in hyperspectral classification, particularly the Graph Neural Network, which exhibits superior capabilities in structural description. However, a single graph structure is not suitable for hyperspectral feature r...

Full description

Saved in:
Bibliographic Details
Main Authors: Chen Wang, Lu Li, ZhongQi Wang, JingYao Ma, YunLong Kong, YanFeng Wang, JianRui Chang, ZiMeng Zhang, XinYu Lin
Format: Article
Language:English
Published: Taylor & Francis Group 2024-12-01
Series:Canadian Journal of Remote Sensing
Online Access:http://dx.doi.org/10.1080/07038992.2024.2333424
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846095213851836416
author Chen Wang
Lu Li
ZhongQi Wang
JingYao Ma
YunLong Kong
YanFeng Wang
JianRui Chang
ZiMeng Zhang
XinYu Lin
author_facet Chen Wang
Lu Li
ZhongQi Wang
JingYao Ma
YunLong Kong
YanFeng Wang
JianRui Chang
ZiMeng Zhang
XinYu Lin
author_sort Chen Wang
collection DOAJ
description In recent years, numerous deep learning-based methods have gained increasing attention in hyperspectral classification, particularly the Graph Neural Network, which exhibits superior capabilities in structural description. However, a single graph structure is not suitable for hyperspectral feature representation. Therefore, we propose a novel Multiple-Scale graph network structure, known as the Multi-Scale Dense Graph Attention network for hyperspectral classification. Firstly, semi-supervised local Fisher discriminant analysis and superpixel segmentation were employed for dimensionality reduction and multi-scale graph construction, respectively. Secondly, Spectral-Spatial convolution is applied to extract shallow features from the image. Subsequently, an improved graph self-attention network is sequentially applied to each scale graph, and the different scale graphs are densely connected through spatial feature alignment modules, designed using twice matrix multiplication. Finally, the combined pixel-level feature map from multiple graph spaces is derived, and Spectral-Spatial convolution is employed to fuse the abundant feature maps for hyperspectral classification. Experimental results on various hyperspectral datasets demonstrate the superiority of our MSDesGATnet over many state-of-the-art methods. The code is available at https://github.com/l7170/MSDesGAT.git.
format Article
id doaj-art-f74960e0bb0548b1bc674aa074652f06
institution Kabale University
issn 1712-7971
language English
publishDate 2024-12-01
publisher Taylor & Francis Group
record_format Article
series Canadian Journal of Remote Sensing
spelling doaj-art-f74960e0bb0548b1bc674aa074652f062025-01-02T11:34:20ZengTaylor & Francis GroupCanadian Journal of Remote Sensing1712-79712024-12-0150110.1080/07038992.2024.23334242333424Multi-Scale Dense Graph Attention Network for Hyperspectral ClassificationChen Wang0Lu Li1ZhongQi Wang2JingYao Ma3YunLong Kong4YanFeng Wang5JianRui Chang6ZiMeng Zhang7XinYu Lin8Department of Artificial Intelligence, School of Automation, Beijing Information Science and Technology UniversityDepartment of Artificial Intelligence, School of Automation, Beijing Information Science and Technology UniversityDepartment of Artificial Intelligence, School of Automation, Beijing Information Science and Technology UniversityDepartment of Artificial Intelligence, School of Automation, Beijing Information Science and Technology UniversityAerospace Information Research Institute, Chinese Academy of SciencesAerospace Information Research Institute, Chinese Academy of SciencesDepartment of Artificial Intelligence, School of Automation, Beijing Information Science and Technology UniversityDepartment of Artificial Intelligence, School of Automation, Beijing Information Science and Technology UniversityDepartment of Artificial Intelligence, School of Automation, Beijing Information Science and Technology UniversityIn recent years, numerous deep learning-based methods have gained increasing attention in hyperspectral classification, particularly the Graph Neural Network, which exhibits superior capabilities in structural description. However, a single graph structure is not suitable for hyperspectral feature representation. Therefore, we propose a novel Multiple-Scale graph network structure, known as the Multi-Scale Dense Graph Attention network for hyperspectral classification. Firstly, semi-supervised local Fisher discriminant analysis and superpixel segmentation were employed for dimensionality reduction and multi-scale graph construction, respectively. Secondly, Spectral-Spatial convolution is applied to extract shallow features from the image. Subsequently, an improved graph self-attention network is sequentially applied to each scale graph, and the different scale graphs are densely connected through spatial feature alignment modules, designed using twice matrix multiplication. Finally, the combined pixel-level feature map from multiple graph spaces is derived, and Spectral-Spatial convolution is employed to fuse the abundant feature maps for hyperspectral classification. Experimental results on various hyperspectral datasets demonstrate the superiority of our MSDesGATnet over many state-of-the-art methods. The code is available at https://github.com/l7170/MSDesGAT.git.http://dx.doi.org/10.1080/07038992.2024.2333424
spellingShingle Chen Wang
Lu Li
ZhongQi Wang
JingYao Ma
YunLong Kong
YanFeng Wang
JianRui Chang
ZiMeng Zhang
XinYu Lin
Multi-Scale Dense Graph Attention Network for Hyperspectral Classification
Canadian Journal of Remote Sensing
title Multi-Scale Dense Graph Attention Network for Hyperspectral Classification
title_full Multi-Scale Dense Graph Attention Network for Hyperspectral Classification
title_fullStr Multi-Scale Dense Graph Attention Network for Hyperspectral Classification
title_full_unstemmed Multi-Scale Dense Graph Attention Network for Hyperspectral Classification
title_short Multi-Scale Dense Graph Attention Network for Hyperspectral Classification
title_sort multi scale dense graph attention network for hyperspectral classification
url http://dx.doi.org/10.1080/07038992.2024.2333424
work_keys_str_mv AT chenwang multiscaledensegraphattentionnetworkforhyperspectralclassification
AT luli multiscaledensegraphattentionnetworkforhyperspectralclassification
AT zhongqiwang multiscaledensegraphattentionnetworkforhyperspectralclassification
AT jingyaoma multiscaledensegraphattentionnetworkforhyperspectralclassification
AT yunlongkong multiscaledensegraphattentionnetworkforhyperspectralclassification
AT yanfengwang multiscaledensegraphattentionnetworkforhyperspectralclassification
AT jianruichang multiscaledensegraphattentionnetworkforhyperspectralclassification
AT zimengzhang multiscaledensegraphattentionnetworkforhyperspectralclassification
AT xinyulin multiscaledensegraphattentionnetworkforhyperspectralclassification