A Simplified 2D-3D CNN Architecture for Hyperspectral Image Classification Based on Spatial–Spectral Fusion
Convolutional neural networks (CNN) have led to a successful breakthrough for hyperspectral image classification (HSIC). Due to the intrinsic spatial-spectral specificities of a hyperspectral cube, feature extraction with 3-D convolution operation is a straightforward way for HSIC. However, the over...
Main Authors: | , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
IEEE
2020-01-01
|
Series: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
Subjects: | |
Online Access: | https://ieeexplore.ieee.org/document/9078778/ |
id |
doaj-dac13086e68c496b8173e38df0654415 |
---|---|
record_format |
Article |
spelling |
doaj-dac13086e68c496b8173e38df06544152021-06-03T23:02:22ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing2151-15352020-01-01132485250110.1109/JSTARS.2020.29832249078778A Simplified 2D-3D CNN Architecture for Hyperspectral Image Classification Based on Spatial–Spectral FusionChunyan Yu0https://orcid.org/0000-0002-9260-6629Rui Han1Meiping Song2https://orcid.org/0000-0002-4489-5470Caiyu Liu3Chein-I Chang4https://orcid.org/0000-0002-5450-4891Information and Technology College, Center of Hyperspectral Imaging in Remote Sensing (CHIRS), Dalian Maritime University, Dalian, ChinaInformation and Technology College, Center of Hyperspectral Imaging in Remote Sensing (CHIRS), Dalian Maritime University, Dalian, ChinaInformation and Technology College, Center of Hyperspectral Imaging in Remote Sensing (CHIRS), Dalian Maritime University, Dalian, ChinaInformation and Technology College, Center of Hyperspectral Imaging in Remote Sensing (CHIRS), Dalian Maritime University, Dalian, ChinaNational Yunlin University of Science and Technology, Yunlin, TaiwanConvolutional neural networks (CNN) have led to a successful breakthrough for hyperspectral image classification (HSIC). Due to the intrinsic spatial-spectral specificities of a hyperspectral cube, feature extraction with 3-D convolution operation is a straightforward way for HSIC. However, the overwhelming features obtained from the original 3-D CNN network suffers from the overfitting and more training cost problem. To address this issue, in this article, a novel HSIC framework based on a simplified 2D-3D CNN is implemented by the cooperation between a 2-D CNN and a 3-D convolution layer. First, the 2-D convolution block aims to extract the spatial features abundantly involved spectral information as a training channel. Then, the 3-D CNN approach primarily concentrates on exploiting band co-relation data by using a reduced kernel. The proposed architecture achieves the spatial and spectral features simultaneously based on a joint 2D-3D pattern to achieve superior fused feature for the subsequent classification. Furthermore, a deconvolution layer intends to enhance the robustness of the deep features is utilized in the proposed CNN network. The results and analysis of extensive real HSIC experiments demonstrate that the proposed light-weighted 2D-3D CNN network can effectively extract refined features and improve the classification accuracy.https://ieeexplore.ieee.org/document/9078778/Convolutionconvolutional neural networks (CNN)feature extractionhyperspectral image classification (HSIC) |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Chunyan Yu Rui Han Meiping Song Caiyu Liu Chein-I Chang |
spellingShingle |
Chunyan Yu Rui Han Meiping Song Caiyu Liu Chein-I Chang A Simplified 2D-3D CNN Architecture for Hyperspectral Image Classification Based on Spatial–Spectral Fusion IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Convolution convolutional neural networks (CNN) feature extraction hyperspectral image classification (HSIC) |
author_facet |
Chunyan Yu Rui Han Meiping Song Caiyu Liu Chein-I Chang |
author_sort |
Chunyan Yu |
title |
A Simplified 2D-3D CNN Architecture for Hyperspectral Image Classification Based on Spatial–Spectral Fusion |
title_short |
A Simplified 2D-3D CNN Architecture for Hyperspectral Image Classification Based on Spatial–Spectral Fusion |
title_full |
A Simplified 2D-3D CNN Architecture for Hyperspectral Image Classification Based on Spatial–Spectral Fusion |
title_fullStr |
A Simplified 2D-3D CNN Architecture for Hyperspectral Image Classification Based on Spatial–Spectral Fusion |
title_full_unstemmed |
A Simplified 2D-3D CNN Architecture for Hyperspectral Image Classification Based on Spatial–Spectral Fusion |
title_sort |
simplified 2d-3d cnn architecture for hyperspectral image classification based on spatial–spectral fusion |
publisher |
IEEE |
series |
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
issn |
2151-1535 |
publishDate |
2020-01-01 |
description |
Convolutional neural networks (CNN) have led to a successful breakthrough for hyperspectral image classification (HSIC). Due to the intrinsic spatial-spectral specificities of a hyperspectral cube, feature extraction with 3-D convolution operation is a straightforward way for HSIC. However, the overwhelming features obtained from the original 3-D CNN network suffers from the overfitting and more training cost problem. To address this issue, in this article, a novel HSIC framework based on a simplified 2D-3D CNN is implemented by the cooperation between a 2-D CNN and a 3-D convolution layer. First, the 2-D convolution block aims to extract the spatial features abundantly involved spectral information as a training channel. Then, the 3-D CNN approach primarily concentrates on exploiting band co-relation data by using a reduced kernel. The proposed architecture achieves the spatial and spectral features simultaneously based on a joint 2D-3D pattern to achieve superior fused feature for the subsequent classification. Furthermore, a deconvolution layer intends to enhance the robustness of the deep features is utilized in the proposed CNN network. The results and analysis of extensive real HSIC experiments demonstrate that the proposed light-weighted 2D-3D CNN network can effectively extract refined features and improve the classification accuracy. |
topic |
Convolution convolutional neural networks (CNN) feature extraction hyperspectral image classification (HSIC) |
url |
https://ieeexplore.ieee.org/document/9078778/ |
work_keys_str_mv |
AT chunyanyu asimplified2d3dcnnarchitectureforhyperspectralimageclassificationbasedonspatialx2013spectralfusion AT ruihan asimplified2d3dcnnarchitectureforhyperspectralimageclassificationbasedonspatialx2013spectralfusion AT meipingsong asimplified2d3dcnnarchitectureforhyperspectralimageclassificationbasedonspatialx2013spectralfusion AT caiyuliu asimplified2d3dcnnarchitectureforhyperspectralimageclassificationbasedonspatialx2013spectralfusion AT cheinichang asimplified2d3dcnnarchitectureforhyperspectralimageclassificationbasedonspatialx2013spectralfusion AT chunyanyu simplified2d3dcnnarchitectureforhyperspectralimageclassificationbasedonspatialx2013spectralfusion AT ruihan simplified2d3dcnnarchitectureforhyperspectralimageclassificationbasedonspatialx2013spectralfusion AT meipingsong simplified2d3dcnnarchitectureforhyperspectralimageclassificationbasedonspatialx2013spectralfusion AT caiyuliu simplified2d3dcnnarchitectureforhyperspectralimageclassificationbasedonspatialx2013spectralfusion AT cheinichang simplified2d3dcnnarchitectureforhyperspectralimageclassificationbasedonspatialx2013spectralfusion |
_version_ |
1721398835545112576 |