混合深度CNN联合注意力的高光谱图像分类

doi:10.3778/j.issn.1673-9418.2104092

摘要/Abstract

摘要： 深度学习中的卷积神经网络（CNN）能充分利用计算机的计算能力，高效地提取遥感图像的特征，取得很好的成果，特别是在高光谱图像分类方面取得了很大的进展。为了在有限的高光谱样本上充分提取光谱和空间特征，提高高光谱图像分类的精度，提出了混合深度卷积联合注意力（HDC-Attention）的模型。首先利用核主成分分析（KPCA）和小批量[K]均值（MBK-means）对高光谱图像进行组合降维，有效地消除数据冗余并保留主要信息量，使得降维后的数据具有最佳区分度。然后将降维后的数据输入HDC网络进行充分的光谱-空间特征提取。最后利用光谱-空间注意力，重新分配光谱-空间特征的权重，增强有用的空谱特征，抑制无用的特征。提出的模型在三个公开数据集上进行了多次实验，在有限的标记样本下，三个数据集的OA、AA、Kappa分类指标均超过99%。

关键词: 高光谱图像分类, 核主成分分析（KPCA）, 卷积神经网络（CNN）, 光谱-空间注意力机制, 深度学习

Abstract: Convolutional neural networks (CNN) in deep learning can make full use of the computing power of computers to efficiently extract the features of remote sensing images. This has achieved good results, especially in the classification of hyperspectral images. In order to fully extract spectral and spatial features from limited hyper spectral samples and improve the accuracy of hyperspectral image classification, a hybrid deep CNN-Attention (HDC-Attention) model is proposed. Firstly, this paper uses kernel principal component analysis (KPCA) and minibatch K-means (MBK-means) to reduce the dimensionality of hyperspectral images. This effectively eliminates data redundancy and retains the main amount of information. The result is that the dimensionality-reduced data have the best discrimination. Then, this paper uses the HDC network to extract the spectral-spatial features from the dimensionality-reduced data. Finally, the spectral-spatial attention is used to redistribute the weights of spectral-spatial features. It enhances useful spatial-spectral features and suppresses useless features. The proposed model has been tested on three public datasets for many times. With a limited number of labeled samples, the OA, AA, and Kappa classification indicators of the three datasets all exceed 99%.

Key words: hyperspectral image classification, kernel principal component analysis (KPCA), convolutional neural networks (CNN), spectral-spatial attention mechanism, deep learning

王燕, 吕艳萍. 混合深度CNN联合注意力的高光谱图像分类[J]. 计算机科学与探索, 2023, 17(2): 385-395.

WANG Yan, LYU Yanping. Hybrid Deep CNN-Attention for Hyperspectral Image Classification[J]. Journal of Frontiers of Computer Science and Technology, 2023, 17(2): 385-395.

参考文献

[1] IMANI M, GHASSEMIAN H. An overview on spectral and spatial information fusion for hyperspectral image classifi-cation: current trends and challenges[J]. Information Fusion, 2020, 59: 59-83.
[2] NAOTO Y, JONATHAN C, KARL S. Potential of resolution-enhanced hyperspectral data for mineral mapping using simulated EnMAP and Sentinel-2 images[J]. Remote Sensing, 2016, 8(3): 172.
[3] LIANG L, DI L, ZHANG L, et al. Estimation of crop LAI using hyperspectral vegetation indices and a hybrid inver-sion method[J]. Remote Sensing of Environment, 2015, 165: 123-134.
[4] 谭海峰, 罗天文, 杨桄, 等. 高光谱异常检测中背景抑制方法研究[J]. 光电子·激光, 2016, 27(2): 177-181.
TAN H F, LUO T W, YANG G, et al. Research on back-ground suppression methods in hyperspectral anomaly detec-tion[J]. Journal of Optoelectronics·Laser, 2016, 27(2): 177-181.
[5] CAPORASO N, WHITWORTH M B, GREBBY S, et al. Non-destructive analysis of sucrose, caffeine and trigonel-line on single green coffee beans by hyperspectral imaging[J]. Food Research International, 2018, 106: 193-203.
[6] ZHANG X, WANG Y, ZHANG N, et al. SSDANet: spectral-spatial three-dimensional convolutional neural network for hyperspectral image classification[J]. IEEE Access, 2020, 8: 127167-127180.
[7] LAI H, DENG J, WEN S. Application of ToF-SIMS and PCA to study interaction mechanism of dodecylamine and smithsonite[J]. Applied Surface Science, 2019, 496: 143698.
[8] HAJER N, RAOUDHA B D, IKRAM AMOUS B A. Effi-cient cloud service discovery approach based on LDA topic modeling[J]. Journal of Systems and Software, 2018, 146: 233-248.
[9] LI Z M, ZHANG J, HUANG H, et al. Semi-supervised bun-dle manifold learning for hyperspectral image classification[J]. Optics and Precision Engineering, 2015, 23(5): 1434-1442.
[10] HE L, LI J, PLAZA A, et al. Discriminative low-rank Gabor filtering for spectral-spatial hyperspectral image classifica-tion[J]. IEEE Transactions on Geoscience and Remote Sen-sing, 2017, 55(3): 1381-1395.
[11] HONG D, WU X, GHAMISI P, et al. Invariant attribute profiles: a spatial-frequency joint feature extractor for hyper-spectral image classification[J]. IEEE Transactions on Geo-science and Remote Sensing, 2020, 58(6): 3791-3808.
[12] CAO X, XU L, MENG D, et al. Integration of 3-dimen-sional discrete wavelet transform and Markov random field for hyperspectral image classification[J]. Neurocomputing, 2017, 226(22): 90-100.
[13] ?ZDEMIR O B, GEDIK E, ?ETIN Y Y. Hyperspectral classification using stacked autoencoders with deep learning[C]//Proceedings of the 6th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing, Lausanne, Jun 24-27, 2014. Piscataway: IEEE, 2014: 1-4.
[14] ZHAO X, CHEN Y S, JIA X P. Spectral-spatial classifica-tion of hyperspectral data based on deep belief network[J]. IEEE Journal of Selected Topics in Applied Earth Observa-tions and Remote Sensing, 2015, 8(6): 2381-2392.
[15] HU W, HUANG Y Y, LI W, et al. Deep convolutional neural networks for hyperspectral image classification[J]. Journal of Sensors, 2015: 258619.
[16] YAO W, LIAN C, BRUZZONE L. ClusterCNN: clustering-based feature learning for hyperspectral image classification[J]. IEEE Geoscience and Remote Sensing Letters,?2021, 18(11): 1991-1995.
[17] 魏祥坡, 余旭初, 谭熊, 等. CNN和三维Gabor滤波器的高光谱图像分类[J]. 计算机辅助设计与图形学学报, 2020, 32(1): 90-98.
WEI X P, YU X C, TAN X, et al. Convolutional neural networks and 3D Gabor filtering for hyperspectral image classification[J]. Journal of Computer-Aided Design & Com-puter Graphics, 2020, 32(1): 90-98.
[18] HAMIDA A B, BENOIT A, LAMBERT P, et al. 3-D deep learning approach for remote sensing image classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 56(8): 4420-4434.
[19] ZHANG C J, LI G D, DU S H. Multi-scale dense networks for hyperspectral remote sensing image classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(11): 9201-9222.
[20] LU Z, XU B, SUN L, et al. 3D channel and spatial attention based multi-scale spatial spectral residual network for hyper-spectral image classification[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2020, 13: 4311-4324.
[21] SUN H, ZHENG X T, LU X Q, et al. Spectral-spatial atten-tion network for hyperspectral image classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020, 58(5): 3232-3245.
[22] FANG B, LI Y, ZHANG H, et al. Hyperspectral images clas-sification based on dense convolutional networks with spectral-wise attention mechanism[J]. Remote Sensing, 2019, 11(2): 159.
[23] LAHDHIRI H, TAOUALI O. Reduced rank KPCA based on GLRT chart for sensor fault detection in nonlinear che-mical process[J]. Measurement, 2021, 169: 108342.
[24] 修瑛昌, 杨文静. Mini Batch K-means算法在遥感影像分类中的应用[J]. 鲁东大学学报(自然科学版), 2017, 33(4): 359-363.
XIU Y C, YANG W J. Application of Mini Batch K-means algorithm in remote sensing data classification[J]. Ludong University Journal (Natural Science Edition), 2017, 33(4): 359-363.
[25] ROY S K, KRISHNA G, DUBEY S R, et al. HybridSN: exploring 3D-2D CNN feature hierarchy for hyperspectral image classification[J]. IEEE Geoscience and Remote Sensing Letters, 2020, 17(2): 277-281.
[26] ZHONG Z L, LI J, LUO Z M, at al. Spectral-spatial resi-dual network for hyperspectral image classification: a 3-D deep learning framework[J]. IEEE Transactions on Geosci-ence and Remote Sensing, 2018, 56(2): 847-858.
[27] MUHAMMAD A. A fast 3D CNN for hyperspectral image classification[J]. IEEE Geoscience and Remote Sensing Let-ters, 2020, 19: 5502205.