计算机科学

计算机科学 ›› 2023, Vol. 50 ›› Issue (6A): 220100158-6.doi: 10.11896/jsjkx.220100158

• 图像处理&多媒体技术 • 上一篇    下一篇

基于改进的ResNeXt网络结构的遥感图像分类

杨星1, 宋玲玲1, 王时绘1,2   

  1. 1 湖北大学计算机与信息工程学院 武汉 430062;
    2 湖北省教育信息化工程技术研究中心 武汉 430062
  • 出版日期:2023-06-10 发布日期:2023-06-12
  • 通讯作者: 王时绘(wsh@hubu.edu.cn)
  • 作者简介:(yxjasson@163.com)
  • 基金资助:
    国家自然科学基金(61902114);湖北省重点实验室开放基金(2020SDSJ06)

Remote Sensing Image Classification Based on Improved ResNeXt Network Structure

YANG Xing1, SONG Lingling1, WANG Shihui1,2   

  1. 1 School of Computer and Information Engineering,Hubei University,Wuhan 430062,China;
    2 Hubei Province Educational Information Engineering Technology Research Center,Wuhan 430062,China
  • Online:2023-06-10 Published:2023-06-12
  • About author:YANG Xing,born in 1997,postgra-duate.His main research interests include deep learning and remote sensing image processing. WANG Shihui,born in 1965,Ph.D,professor,master supervisor,is a member of China Computer Federation.His main research interests include machine learning and software engineering.
  • Supported by:
    National Natural Science Foundation of China(61902114) and Hubei Provincial Key Laboratory Open Fund(2020SDSJ06).

PDF (PC)

327

摘要： 遥感图像分类是遥感图像信息处理的关键方向之一,其分类精准率很大程度上限制了遥感技术整体的发展。对于遥感图像,传统机器学习算法与模型结构存在不能快速提取特征图,且分类结果不够准确的缺陷。针对这一问题,提出了一种改进的基于ResNeXt网络模型结合注意力机制,以优化后SVM(支持向量机)算法替换全连接层的模型。首先引入计算机视觉中的注意力机制,对不同特征赋予不同的权重,提高对图像中用于分类部分有效信息的提取能力,然后结合ResNeXt网络,最后以优化后的SVM算法替换卷积神经网络末端的全连接层用于提升分类效果,同时在模型整体不增加超参数的情况下优化了网络性能。该网络模型在数据集AID上的实验结果表明,改进后的网络模型对深层特征的提取能力有显著提高,且优化后的网络模型对于多分类任务具有较优的分类效果。

关键词: 遥感图像, 卷积神经网络, ResNeXt, 注意力机制, 场景分类, SVM

Abstract: Remote sensing image classification is one of the key directions of remote sensing image information processing,and its classification accuracy greatly limits the overall development of remote sensing technology.Traditional machine learning algorithms and model structures have the disadvantages that they cannot quickly extract feature maps from remote sensing images,and the classification results are not accurate enough.Aiming at this problem,an improved model based on the ResNeXt network model combined with the attention mechanism is proposed to replace the fully connected layer model with the optimized SVM(support vector machine) algorithm.Firstly,it introduces the attention mechanism in computer vision,assigns different weights to different features,improves the ability to extract effective information for the classification part of the image,then combines the ResNeXt network,and finally replaces the end of the convolutional neural network with the optimized SVM algorithm.The fully connected layer is used to improve the classification effect,and at the same time optimize the network performance without increasing the hyperparameters of the model as a whole.Experimental results of the network model on the data set AID show that the improved network model has a significant improvement in the ability to extract deep features,and the optimized network mo-del has a better classification effect for multi-classification tasks.

Key words: Remote sensing image, CNN, ResNeXt, Attention mechanism, Scene classification, SVM

中图分类号: 

  • TP751
[1]HASHEMI B L,GEBREHIWOT A.Deep learning for remotesensing image classification for agriculture applications[J].ISPRS-International Archives of the Photogrammetry,Remote Sensing and Spatial Information Sciences,2020,44:51-54.
[2]LU D L,NING Q,ZANG J.Improved KNN algorithm based on BP neural network decision-making[J].Computer Applications,2017,37(S2):65-67,88.
[3]SPINELLA M.Improved principal component analysis and li-near regression classification for face recognition[J].Signal Processing,2018,145:175-182.
[4]FU W F,ZOU W B.Research progress of deep learning in remote sensing image classification[J].Application Research of Computers,2018,326(12):3521-3525.
[5]BU X L,HUO H,FANG T.Rotating r-obust remote sensingimage feature extraction based on sparse representation[J].Computer Engineering,2012,38409(14):124-127.
[6]NIU X X,SUN A,WANG Y F,et al.Research on remote sen-sing image classification based on deep learning[J].Laser Journal,2021,42(5):10-14.
[7]GUO Y Y,YU J,DU X S,et al.Outlier detection algorithm based on autoencoder and integrated learning[J].Computer Applications,2022,42(1):2018.
[8]VADDI R,MANOHARAN P.Hyperspectral remote sensingimage classification using combinatorial optimisation based un-supervised band selection and CNN[J].IET Image Processing,2020,14(15):3909-3919.
[9]ZHAO C J,ZHOU S G,DING Q,et al.Semi-supervised classification of hyperspectral images based on homogeneity area and transfer learning[J].World of Geographic Information,2019,26(5):45-52.
[10]WAN Y L,ZHONG X W,LIU H,et al.A review of applications of convolutional neural networks in hyperspectral image classification[J].Computer Engineering and Applications,2021,57(4):1-10.
[11]LI W,FAN Y C,JIANG Q Y,et al.Medical image classification method based on variable convolutional autoencoder optimized for teaching and learning[J].Computer Applications,2022,42(2):592-598.
[12]JIANG P Y,TAO Q C,AI M Q.Image Classification of Steel Surface Defects Based on Attention Mechanism and Deep Lear-ning[J].Computer Applications and Software,2021,38(9):214-219.
[13]CHENG D Q,WANG Y C,KOU Q Q,et al.Mine Image Classification Based on Improved Deep Residual Network[J].Computer Application Research,2021,38(5):1576-1580.
[14]WANG X,ZHANG H Y,LING C.Semantic Segmentation ofSAR Remote Sensing Image Based on U-Net Optimization[J].Computer Science,2021,48(11A):376-381.
[15]YUAN X X,WU Q.Object Detection in Remote Sensing Images Based on Saliency Feature and Angle Information[J].Computer Science,2021,48(4):174-179.
[16]LIU J W,LIU J W,LUO X L.Research progress of attention mechanism in deep learning[J].Journal of Engineering Science,2021,43(11):1499-1511.
[17]ZHANG B,ZHANG X J,ZHAO B C,et al.Application of animproved DenseASPP network in remote sensing image segmentation[J].Computer Applications and Software,2021,38(7):46-52.
[18]QIAO X X,SHI W Z,LIU X X,et al.Remote sensing imagescene classification based on ResNet dual attention mechanism[J].Computer System Applications,2021,30(8):243-248.
[19]LI W S,WANG Z X,LI S H,et al.Weakly supervised fine-grained image classification based on attention mechanism[J].Computer Systems Applications,2021,30(10):232-239.
[20]CHENG G,HAN J W,LU X Q.Remote Sensing Image Scene Classification:Benchmark and State of the Art[J].Proceedings of the IEEE, 2017,105(10):1865-1883.
[21]XIA G S,HU J,HU F,et al.AID:A Benchmark Dataset forPerformance Evaluation of Aerial Scene Classification[J].IEEE Transactions on Geoscience and Remote Sensing,2017,55(7):3965-3981.
[22]ZHANG M L,ZHOU Z H.ML-KNN:A lazy learning approach to multi-label leaming[J].Pattern Recognition:The Journal of the Pattern Recognition Society,2007,40(7):2038-2048.
[23]NIR F,DAN G,MOISES G.Bayesian Network Classifiers.[J].Machine Learning,1997,29(2/3):131-163.
[24]SUN H,CHEN J,LEI L,et al.Summarization of Anti-robustness Technology of Deep Convolutional Neural Network Image Recognition Model[J].Journal of Radars,2021,10(4):571-594.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市渝北区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发