: .
第 44 卷 第 1 期 imensionality of a hyperspectral dataset. Second, the principal components obtained through
dimensionality reduction are used to generate a multidimensional feature set through super-pixel segmentation
and cosine clustering. Finally, the superimposed feature set is used to extract spatial-spectrum features through
a two-dimensional and three-dimensional multi-scale hybrid convolutional network, and a capsule network is
used to classify them. We performed experiments on different hyperspectral datasets, and the results revealed
that under the same 20-dimensional spectral setting, the proposed strategy significantly improves the overall
accuracy, average accuracy, and Kappa coefficient compared to traditional classification strategies.
Key words: image classification, hyperspectral image, dimensionality reduction, superpixel, hybrid
convolution capsules network
谱图像(hyperspectral image,HSI)能够捕捉复杂的地
0 引言
物分布,因此 HSI 被广泛应用于深林植被保护、水质
高光谱遥感是指在电磁波谱的紫外、可见光、近 检测和地质勘测等[2]。然而,由于大气分子影响使图
红外和中红外区域,利用成像光谱仪获取窄而连续的 像存在椒盐噪声,以及光谱高维性所带来的冗余,给
光谱图像数据[1]。由于包含丰富的光谱信息使得高光 遥感图像实现准确的分类带来了极大挑战。
收稿日期:2020-11-02;修订日期:2021-01-25.
作者简介:熊余(1982-),男,研究员,博士,主要研究方向为教育大数据,光网络。E-mail:******@。
基金项目:国家自然科学基金资助项目(61401052); 国家留学基金委资助项目(201608500030); 重庆市教委科学技术研究资助项目(KJ1400418,
KJ1500445); 重庆邮电大学博士启动基金资助项目(A2015-09)。
9第 44 卷 第 1 期 红 外 技 术
2022 年 1 月 Infrared Technology Jan
多特征融合下的高光谱图像混合卷积分类 来自淘豆网m.daumloan.com转载请标明出处.