Hyperspectral image classification using 3D-2D CNN with multi-scale information extraction and fusion module

Hang Gong, Tingkui Mu, Qiuxia Li, Feng Han, Abudusalamu Tuniyazi, Haoyang Li, Wenjing Wang, Zhiping He, Chunlai Li, H. Dai
{"title":"Hyperspectral image classification using 3D-2D CNN with multi-scale information extraction and fusion module","authors":"Hang Gong, Tingkui Mu, Qiuxia Li, Feng Han, Abudusalamu Tuniyazi, Haoyang Li, Wenjing Wang, Zhiping He, Chunlai Li, H. Dai","doi":"10.1117/12.2586803","DOIUrl":null,"url":null,"abstract":"Classification is the focus and difficulty of hyperspectral imaging technology. Hyperspectral data have twodimensional spatial information and one-dimensional spectral information, which are presented as three-dimensional data blocks with large amount of information, meanwhile high-dimension, high nonlinearity and limited training samples bring great challenges. Deep learning can extract and analyze the features of target data step by step by building multi-layer deep nonlinear structure. The advanced feature, multi scale abstract information extracted by convolution neural network applied to image processing can improve the classification accuracy of complex hyperspectral data. We regard pixel level hyperspectral classification as semantic segmentation network, and creatively introduce squeeze-and-excitation network and pyramid pooling network into hyperspectral classification network and proposed a model based on the structure of 2D-3D hybrid convolution neural network, it can learn deeper spatial spectral features and fusion to improve the accuracy and speed of hyperspectral classification.","PeriodicalId":370739,"journal":{"name":"International Conference on Photonics and Optical Engineering and the Annual West China Photonics Conference","volume":"15 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2021-01-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"International Conference on Photonics and Optical Engineering and the Annual West China Photonics Conference","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1117/12.2586803","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 1

Abstract

Classification is the focus and difficulty of hyperspectral imaging technology. Hyperspectral data have twodimensional spatial information and one-dimensional spectral information, which are presented as three-dimensional data blocks with large amount of information, meanwhile high-dimension, high nonlinearity and limited training samples bring great challenges. Deep learning can extract and analyze the features of target data step by step by building multi-layer deep nonlinear structure. The advanced feature, multi scale abstract information extracted by convolution neural network applied to image processing can improve the classification accuracy of complex hyperspectral data. We regard pixel level hyperspectral classification as semantic segmentation network, and creatively introduce squeeze-and-excitation network and pyramid pooling network into hyperspectral classification network and proposed a model based on the structure of 2D-3D hybrid convolution neural network, it can learn deeper spatial spectral features and fusion to improve the accuracy and speed of hyperspectral classification.
采用3D-2D CNN多尺度信息提取融合模块进行高光谱图像分类
分类是高光谱成像技术的重点和难点。高光谱数据具有二维空间信息和一维光谱信息,以三维数据块的形式呈现,信息量大,同时高维、高非线性和有限的训练样本给高光谱数据处理带来很大挑战。深度学习可以通过构建多层深度非线性结构,逐步提取和分析目标数据的特征。将卷积神经网络提取的先进特征、多尺度抽象信息应用于图像处理,可以提高复杂高光谱数据的分类精度。我们将像素级高光谱分类作为语义分割网络,创造性地将挤压激励网络和金字塔池化网络引入到高光谱分类网络中,提出了一种基于2D-3D混合卷积神经网络结构的模型,该模型可以学习更深层次的空间光谱特征并进行融合,从而提高高光谱分类的精度和速度。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:604180095
Book学术官方微信