基于卷积增强动态图卷积特征提取的高光谱图像分类  

Hyperspectral Image Classification Based on Enhanced Dynamic-Graph-Convolutional Feature Extraction

作　　者：李铁[1] 高乔裕 李文许 Li Tie;Gao Qiaoyu;Li Wenxu(School of Electronics and Information Engineering,Liaoning University of Engineering and Techonlogy,Huludao 125105,Liaoning,China)

机构地区：[1]辽宁工程技术大学电子与信息工程学院,辽宁葫芦岛125105

出　　处：《激光与光电子学进展》2024年第18期429-438,共10页Laser & Optoelectronics Progress

基　　金：辽宁省科技厅自然科学基金面上项目(2023-MS-314);辽宁省教育厅科学研究经费项目(LJ2020JCL007);辽宁省教育厅基本科研面上项目(LJKMZ20220678,LJKZ0357)。

摘　　要：针对高光谱图像存在的光谱维度多、数据分布不均匀、空间光谱特征提取不充分、同物异谱和异物同谱等问题,提出一种基于卷积网络结合图神经网络的高光谱图像分类算法。首先,通过主成分分析对高光谱图像进行降维;然后,利用卷积网络提取图像中的局部特征,包括纹理和形状等信息。这些局部特征可以反映图像中不同物体和区域间的差异,同时将卷积所提取到的特征通过编码器嵌入动态图卷积所在的超像素域中,利用其动态化的邻接矩阵来捕获高光谱图像中的长远距离空间上下文信息并通过解码器将二者所提取到的特征相结合,从而实现对高光谱图像中各个像素类别的分类。在3个公共的高光谱图像数据集上进行实验,与其他5种高光谱图像分类算法相比,所提算法具有更好的分类性能。Herein,a hyperspectral image classification algorithm that integrates convolutional network and graph neural network is proposed to address several challenges,such as high spectral dimensionality,uneven data distribution,inadequate spatial-spectral feature extraction,and spectral variability.First,principal component analysis is performed to reduce the dimensionality of hyperspectral images.Subsequently,convolutional networks extract local features,including texture and shape information,highlighting differences between various objects and regions within the image.The extracted features are then embedded into the superpixel domain,where dynamic graph convolution occurs via an encoder.A dynamic adjacency matrix captures the long-term spatial context information in the hyperspectral image.These features are combined through a decoder to effectively classify different pixel categories.Experiments conducted on three commonly used hyperspectral image datasets demonstrate that this method outperforms five other classification techniques with regard to classification performance.

关 键 词：高光谱图像 超像素 卷积神经网络 图神经网络 动态图卷积 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]