Journal of Frontiers of Computer Science and Technology ›› 2023, Vol. 17 ›› Issue (2): 385-395.DOI: 10.3778/j.issn.1673-9418.2104092

• Graphics·Image • Previous Articles     Next Articles

Hybrid Deep CNN-Attention for Hyperspectral Image Classification

WANG Yan, LYU Yanping   

  1. School of Computer and Communication, Lanzhou University of Technology, Lanzhou 730050, China
  • Online:2023-02-01 Published:2023-02-01



  1. 兰州理工大学 计算机与通信学院,兰州 730050

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

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