SIA OpenIR  > 光电信息技术研究室
基于改进残差密集网络的高光谱重建
Alternative TitleHyperspectral Images Reconstruction Based on Improved Residual
李勇1; 金秋雨1,2; 赵怀慈2; 李波3
Department光电信息技术研究室
Source Publication光学学报
ISSN0253-2239
2020
Pages1-15
Contribution Rank2
Funding Organization辽宁省自然科学基金(2019-ZD-0205) ; 辽宁省自然科学基金(2019-MS-238)
Keyword高光谱成像 残差密集网络 通道自适应 特征重标定 RGB图像
Abstract

高光谱图像包含丰富的光谱信息,单幅RGB重建高光谱图像对于军事目标识别和医学诊断领域有重要价值。而传统算法无法对未知相机光谱响应的RGB图像进行重建,针对此问题,提出了一种改进的残差密集网络。将改进的残差密集块作为残差密集网络的基本模块,使用自适应权重模块对特征通道进行特征重标定,使得高光谱重建精度得到了提高。其次,使用特征变换层替代了原来网络的空间变换层,将解决图像超分辨率问题转换成解决高光谱重建问题,实现了网络从空间维度到光谱维度的转变。实验结果证明,本文的方法无论是在主观效果还是客观评估指标上均优于主流的传统方法和深度学习方法,对比稀疏字典方法,平均相对绝对误差(MRAE)和均方误差(RMSE)分别下降了46.7%和44.8%。

Other Abstract

Hyperspectral images contain rich spectral information. The hyperspectral images reconstruction from a single RGB image is of great value in the field of military target recognition and medical diagnosis. Traditional algorithms cannot reconstruct RGB images with unknown camera spectral response. To solve this problem, an improved residual dense network is proposed. Using an improved residual dense block as the basic module of the residual dense network and the feature channels are recalibrated by the auto-adaptive weight module, which improves the accuracy of hyperspectral reconstruction. Additionally, the feature transformation layer is used to replace the spatial transformation layer of the original network, which converts the problem of image super-resolution to the problem of hyperspectral reconstruction, and realizes the transformation of the network from the spatial dimension to the spectral dimension. The experimental results demonstrate that our proposed method is superior to the existing traditional methods and deep learning methods in both subjective effect and objective evaluation indicators. Compared with the sparse dictionary method, the Mean Relative Absolute Error (MRAE) and Root Mean Square Error (RMSE) are reduced by 46.7% and 44.8% respectively.

Language中文
Document Type期刊论文
Identifierhttp://ir.sia.cn/handle/173321/27991
Collection光电信息技术研究室
Corresponding Author赵怀慈
Affiliation1.沈阳工业大学电气工程学院
2.中国科学院沈阳自动化研究所光电信息处理重点实验室
3.沈阳工程学院信息学院
Recommended Citation
GB/T 7714
李勇,金秋雨,赵怀慈,等. 基于改进残差密集网络的高光谱重建[J]. 光学学报,2020:1-15.
APA 李勇,金秋雨,赵怀慈,&李波.(2020).基于改进残差密集网络的高光谱重建.光学学报,1-15.
MLA 李勇,et al."基于改进残差密集网络的高光谱重建".光学学报 (2020):1-15.
Files in This Item:
File Name/Size DocType Version Access License
基于改进残差密集网络的高光谱重建.pdf(1439KB)期刊论文出版稿开放获取CC BY-NC-SAView Application Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[李勇]'s Articles
[金秋雨]'s Articles
[赵怀慈]'s Articles
Baidu academic
Similar articles in Baidu academic
[李勇]'s Articles
[金秋雨]'s Articles
[赵怀慈]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[李勇]'s Articles
[金秋雨]'s Articles
[赵怀慈]'s Articles
Terms of Use
No data!
Social Bookmark/Share
File name: 基于改进残差密集网络的高光谱重建.pdf
Format: Adobe PDF
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.