|Table of Contents|

Multispectral and Panchromatic Remote Sensing Image Fusion AlgorithmBased on Convolutional Neural Networks(PDF)

《南京师大学报（自然科学版）》[ISSN:1001-4616/CN:32-1239/N]

Issue:
2021年03期
Page:
123-130
Research Field:
·计算机科学与技术·
Publishing date:

Info

Title:
Multispectral and Panchromatic Remote Sensing Image Fusion AlgorithmBased on Convolutional Neural Networks
Author(s):
Han Wenjun1; Sun Xiaohu1; Ji Genlin2; Su Xiaoyun3; Xie Fei3; Wu Bing4; Chen Hong5
(1.State Grid Economic and Technology Research Institute Co. Ltd,Beijing 102209,China)(2.School of Computer and Electronic Information,Nanjing Normal University,Nanjing 210023,China)(3.School of Electrical and Automation Engineering,Nanjing Normal University,Nanjing 210023,China)(4.Economic and Technology Research Institute of Zhejiang Electric Power Co. Ltd,Hangzhou 310000,China)(5.Economic and Technology Research Institute of Jiangsu Electric Power Co. Ltd,Nanjing 210008,China)
Keywords:
convolutional neural networks; multispectral image; panchromatic image; fusion algorithm
PACS:
TP751
DOI:
10.3969/j.issn.1001-4616.2021.03.018
Abstract:
The geographic information data has multi-scale characteristics such as different resolution,different precision and different coverage. In application,it is often necessary to have both high spectral resolution and high spatial resolution images. In order to improve the accuracy of multispectral image and panchromatic image fusion,a fusion algorithm for multispectral and panchromatic remote sensing images based on convolutional neural networks is proposed. Firstly,the images in the training set is preprocessed to construct the image data set suitable for the algorithm. Secondly,the convolution layers of convolutional neural networks are expanded to improve the three channel correlation and extract more image information. Finally,the depth separable convolutional neural networks are used to improve the accuracy of the fusion image,which can also accelerate the fusion speed. The accuracy and speed of the algorithm are verified using the original images provided by downsampled ImageNet dataset. Experimental results show that the mean square error of the proposed algorithm is 7% lower and the fusion time is reduced by 29% compared with the traditional convolutional neural networks method,and obtain a good fusion effect.

References:

[1] 李昌洁,宋慧慧,张开华,等. 条件生成对抗遥感图像时空融合[J]. 中国图象图形学报,2021,26(3):714-726.
[2]李树涛,李聪妤,康旭东. 多源遥感图像融合发展现状与未来展望[J]. 遥感学报,2021,25(1):148-166.
[3]黄波,姜晓璐. 增强型空间像元分解时空遥感影像融合算法[J]. 遥感学报,2021,25(1):241-250.
[4]顾宇鑫,马小虎. 采用稀疏变换和拉普拉斯金字塔的数字水印算法[J]. 计算机辅助设计与图形学学报,2018,30(5):901-910.
[5]傅志中,王雪,李晓峰,等. 基于视觉显著性和NSCT的红外与可见光图像融合[J]. 电子科技大学学报,2017,46(2):357-362.
[6]殷明,段普宏,褚标,等. 基于非下采样双树复轮廓波变换和稀疏表示的红外和可见光图像融合[J]. 光学精密工程,2016,24(7):1763-1771.
[7]刘先红,陈志斌,秦梦泽. 结合引导滤波和卷积稀疏表示的红外与可见光图像融合[J]. 光学精密工程,2018,26(5):1242-1253.
[8]张建明,邱晓晖. 基于Curvelet变换的指纹图像去噪[J]. 计算机技术与发展,2018,28(5):164-167.
[9]颜正恕,王璟. 基于非下采样轮廓波变换耦合对比度特征的遥感图像融合算法[J]. 电子测量与仪器学报,2020,34(3):28-35.
[10]GHASSEMIAN H. A review of remote sensing image fusion methods[J]. Information fusion,2016,32:75-89.
[11]FAN C,WANG L,LIU P,et al. Compressed sensing based remote sensing image reconstruction via employing similarities of reference images[J]. Multimedia tools and applications,2016,75(19):12201-12225.
[12]LI C M,ZHANG H G,WU P D,et al. A complex junction recognition method based on GoogLeNet model[J]. Transactions in GIS,2020,24(6):1756-1778.
[13]Srivastava R K,Greff K,Schmidhuber J. Training very deep networks[C]//Proceedings of the Advances in Neural Information Processing Systems. Montreal,Canada:MIT Press,2015:2377-2385.
[14]许可,高尚. 深度卷积神经网络LeNet-5和ResNet的对比以及应用分析[J]. 电子设计工程,2020,28(2):82-85.
[15]马永杰,刘培培. 基于DenseNet进化卷积神经网络的图像分类算法[J]. 激光与光电子学进展,2020,57(24):1-5.
[16]GANGKOFNER U G,PRADHAN P S,HOLCOMB D W. Optimizing the high-pass filter addition technique for image fusion[J]. Photogrammetric engineering & remote sensing,2007,73(9):1107-1118.
[17]李红,刘芳,杨淑媛,等. 基于深度支撑值学习网络的遥感图像融合[J]. 计算机学报,2016,39(8):1583-1596.

Memo

Memo:
-
Last Update: 2021-09-15