Consolidated Convolutional Neural Network for Hyperspectral Image Classification
The performance of hyperspectral image (HSI) classification is highly dependent on spatial and spectral information, and is heavily affected by factors such as data redundancy and insufficient spatial resolution. To overcome these challenges, many convolutional neural networks (CNN) especially 2D-CN...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-03-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/14/7/1571 |
| _version_ | 1797437944204623872 |
|---|---|
| author | Yang-Lang Chang Tan-Hsu Tan Wei-Hong Lee Lena Chang Ying-Nong Chen Kuo-Chin Fan Mohammad Alkhaleefah |
| author_facet | Yang-Lang Chang Tan-Hsu Tan Wei-Hong Lee Lena Chang Ying-Nong Chen Kuo-Chin Fan Mohammad Alkhaleefah |
| author_sort | Yang-Lang Chang |
| collection | DOAJ |
| description | The performance of hyperspectral image (HSI) classification is highly dependent on spatial and spectral information, and is heavily affected by factors such as data redundancy and insufficient spatial resolution. To overcome these challenges, many convolutional neural networks (CNN) especially 2D-CNN-based methods have been proposed for HSI classification. However, these methods produced insufficient results compared to 3D-CNN-based methods. On the other hand, the high computational complexity of the 3D-CNN-based methods is still a major concern that needs to be addressed. Therefore, this study introduces a consolidated convolutional neural network (C-CNN) to overcome the aforementioned issues. The proposed C-CNN is comprised of a three-dimension CNN (3D-CNN) joined with a two-dimension CNN (2D-CNN). The 3D-CNN is used to represent spatial–spectral features from the spectral bands, and the 2D-CNN is used to learn abstract spatial features. Principal component analysis (PCA) was firstly applied to the original HSIs before they are fed to the network to reduce the spectral bands redundancy. Moreover, image augmentation techniques including rotation and flipping have been used to increase the number of training samples and reduce the impact of overfitting. The proposed C-CNN that was trained using the augmented images is named C-CNN-Aug. Additionally, both Dropout and L2 regularization techniques have been used to further reduce the model complexity and prevent overfitting. The experimental results proved that the proposed model can provide the optimal trade-off between accuracy and computational time compared to other related methods using the Indian Pines, Pavia University, and Salinas Scene hyperspectral benchmark datasets. |
| first_indexed | 2024-03-09T11:28:50Z |
| format | Article |
| id | doaj.art-c9745743a99242c2aa4158b3d829fa91 |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-09T11:28:50Z |
| publishDate | 2022-03-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-c9745743a99242c2aa4158b3d829fa912023-11-30T23:55:54ZengMDPI AGRemote Sensing2072-42922022-03-01147157110.3390/rs14071571Consolidated Convolutional Neural Network for Hyperspectral Image ClassificationYang-Lang Chang0Tan-Hsu Tan1Wei-Hong Lee2Lena Chang3Ying-Nong Chen4Kuo-Chin Fan5Mohammad Alkhaleefah6Department of Electrical Engineering, National Taipei University of Technology, Taipei 10608, TaiwanDepartment of Electrical Engineering, National Taipei University of Technology, Taipei 10608, TaiwanDepartment of Electrical Engineering, National Taipei University of Technology, Taipei 10608, TaiwanDepartment of Communications, Navigation and Control Engineering, National Taiwan Ocean University, Keelung City 202301, TaiwanCenter for Space and Remote Sensing Research, National Central University, Taoyuan 32001, TaiwanDepartment of Computer Science & Information Engineering, National Central University, Taoyuan 32001, TaiwanDepartment of Electrical Engineering, National Taipei University of Technology, Taipei 10608, TaiwanThe performance of hyperspectral image (HSI) classification is highly dependent on spatial and spectral information, and is heavily affected by factors such as data redundancy and insufficient spatial resolution. To overcome these challenges, many convolutional neural networks (CNN) especially 2D-CNN-based methods have been proposed for HSI classification. However, these methods produced insufficient results compared to 3D-CNN-based methods. On the other hand, the high computational complexity of the 3D-CNN-based methods is still a major concern that needs to be addressed. Therefore, this study introduces a consolidated convolutional neural network (C-CNN) to overcome the aforementioned issues. The proposed C-CNN is comprised of a three-dimension CNN (3D-CNN) joined with a two-dimension CNN (2D-CNN). The 3D-CNN is used to represent spatial–spectral features from the spectral bands, and the 2D-CNN is used to learn abstract spatial features. Principal component analysis (PCA) was firstly applied to the original HSIs before they are fed to the network to reduce the spectral bands redundancy. Moreover, image augmentation techniques including rotation and flipping have been used to increase the number of training samples and reduce the impact of overfitting. The proposed C-CNN that was trained using the augmented images is named C-CNN-Aug. Additionally, both Dropout and L2 regularization techniques have been used to further reduce the model complexity and prevent overfitting. The experimental results proved that the proposed model can provide the optimal trade-off between accuracy and computational time compared to other related methods using the Indian Pines, Pavia University, and Salinas Scene hyperspectral benchmark datasets.https://www.mdpi.com/2072-4292/14/7/1571consolidated convolutional neural networkhyperspectral image classificationhigh performance computingimage augmentationprincipal component analysis |
| spellingShingle | Yang-Lang Chang Tan-Hsu Tan Wei-Hong Lee Lena Chang Ying-Nong Chen Kuo-Chin Fan Mohammad Alkhaleefah Consolidated Convolutional Neural Network for Hyperspectral Image Classification Remote Sensing consolidated convolutional neural network hyperspectral image classification high performance computing image augmentation principal component analysis |
| title | Consolidated Convolutional Neural Network for Hyperspectral Image Classification |
| title_full | Consolidated Convolutional Neural Network for Hyperspectral Image Classification |
| title_fullStr | Consolidated Convolutional Neural Network for Hyperspectral Image Classification |
| title_full_unstemmed | Consolidated Convolutional Neural Network for Hyperspectral Image Classification |
| title_short | Consolidated Convolutional Neural Network for Hyperspectral Image Classification |
| title_sort | consolidated convolutional neural network for hyperspectral image classification |
| topic | consolidated convolutional neural network hyperspectral image classification high performance computing image augmentation principal component analysis |
| url | https://www.mdpi.com/2072-4292/14/7/1571 |
| work_keys_str_mv | AT yanglangchang consolidatedconvolutionalneuralnetworkforhyperspectralimageclassification AT tanhsutan consolidatedconvolutionalneuralnetworkforhyperspectralimageclassification AT weihonglee consolidatedconvolutionalneuralnetworkforhyperspectralimageclassification AT lenachang consolidatedconvolutionalneuralnetworkforhyperspectralimageclassification AT yingnongchen consolidatedconvolutionalneuralnetworkforhyperspectralimageclassification AT kuochinfan consolidatedconvolutionalneuralnetworkforhyperspectralimageclassification AT mohammadalkhaleefah consolidatedconvolutionalneuralnetworkforhyperspectralimageclassification |