摘要
在高光谱图像分类中,原始的图卷积网络作用在数据量较大的数据集上时,会出现内存开销大、时间成本高的问题,而且单一的图卷积网络模型不能对高光谱图像进行充分的特征提取。为了在数据量较大时降低时间成本并充分提取特征以提高分类精度,本文研究了Gabor滤波和批处理的图卷积网络级联卷积神经网络的融合网络对高光谱图像进行特征提取的方法,并在3个数据集上进行了验证。实验结果表明,本文的方法在对数据量较大的数据集分类时可以较好地降低时间成本,提高分类精度。
In hyperspectral image classification,when original graph convolution network acts on a data set with a large amount of data,there will be problems of large memory overhead and high time cost.Moreover,a single graph convolution network model cannot fully extract the features of hyperspectral images.To reduce the time cost and fully extract features to improve the classification accuracy when the amount of data is large,this paper studies the feature extraction method of hyperspectral images based on the fusion network of Gabor filtering and batch image convolution network cascaded convolution neural network,and verifies it on three data sets.Experimental results show that this method can reduce the time cost and improve the classification accuracy when classifying data sets with a large amount of data.
作者
王婷婷
陈立伟
崔颖
高山
WANG Tingting;CHEN Liwei;CUI Ying;GAO Shan(College of Information and Communications Engineering,Harbin Engineering University,Harbin 150001,China)
出处
《应用科技》
CAS
2023年第2期79-85,共7页
Applied Science and Technology
基金
国家自然科学基金项目(61102105).
关键词
高光谱图像分类
GABOR
图卷积
卷积神经网络
K近邻
特征提取
级联融合
批处理
hyperspectral image classification
Gabor
graph convolution
convolutional neural network
K-nearest neighbor
feature extraction
cascade fusion
batch processing
