Please wait a minute...
 
国土资源遥感  2021, Vol. 33 Issue (1): 45-53    DOI: 10.6046/gtzyyg.2020162
     技术方法 本期目录 | 过刊浏览 | 高级检索 |
基于语义分割网络的高分遥感影像城市建成区提取方法研究与对比分析
刘钊(), 赵桐(), 廖斐凡, 李帅, 李海洋
清华大学土木工程系交通工程与地球空间信息研究所,北京 100084
Research and comparative analysis on urban built-up area extraction methods from high-resolution remote sensing image based on semantic segmentation network
LIU Zhao(), ZHAO Tong(), LIAO Feifan, LI Shuai, LI Haiyang
Institute of Transportation Engineering and Geomatics, Department of Civil Engineering, Tsinghua University, Beijing 100084, China
全文: PDF(4183 KB)   HTML  
输出: BibTeX | EndNote (RIS)      
摘要 

城市建成区的提取对城市发展规划有着重要的作用。为了找出能兼顾效率和识别准确率的基于卷积神经网络的遥感影像城市建成区提取方法,从神经网络结构的原理出发,对多种语义分割网络的内部结构进行对比分析,并针对语义分割网络分别进行训练及结果比较。实验结果表明,ShelfNet-50网络能够在训练速度最快的同时保证很高的识别准确率,在训练时长仅需14 h的同时达到了77%的前景分割精度,且ShelfNet-50网络预测的结果也与相应的遥感影像数据高度吻合。实验说明ShelfNet-50网络可应用于高分遥感影像的城市建成区提取研究。

服务
把本文推荐给朋友
加入引用管理器
E-mail Alert
RSS
作者相关文章
刘钊
赵桐
廖斐凡
李帅
李海洋
关键词 高分遥感影像卷积神经网络语义分割城市建成区    
Abstract

The extraction of urban built-up areas plays an important role in urban development planning. To find out the method of extracting remote sensing image urban built-up area based on convolutional neural network which can balance efficiency and recognition accuracy, the authors started with the principle of neural network structure and compared as well as analyzed the internal structure of multiple semantic segmentation networks. The semantic segmentation network was trained separately and the results were comparatively studied. The experimental result shows that the ShelfNet-50 network could ensure high recognition accuracy while training speed, achieved 77% foreground segmentation accuracy while training time was only 14 hours, and the result of ShelfNet-50 network prediction was also highly consistent with the corresponding remote sensing image data. The experiment confirms that ShelfNet-50 network can be applied to high-resolution remote sensing image urban built-up area extraction problems.

Key wordshigh-resolution remote sensing image    convolutional neural network    semantic segmentation    urban built-up area
收稿日期: 2020-06-09      出版日期: 2021-03-18
ZTFLH:  P237  
通讯作者: 赵桐
作者简介: 刘 钊(1967-),男,副教授,主要从事GIS及其应用、云GIS、时空大数据及遥感图像处理等方面的研究。Email: liuz@tsinghua.edu.cn
引用本文:   
刘钊, 赵桐, 廖斐凡, 李帅, 李海洋. 基于语义分割网络的高分遥感影像城市建成区提取方法研究与对比分析[J]. 国土资源遥感, 2021, 33(1): 45-53.
LIU Zhao, ZHAO Tong, LIAO Feifan, LI Shuai, LI Haiyang. Research and comparative analysis on urban built-up area extraction methods from high-resolution remote sensing image based on semantic segmentation network. Remote Sensing for Land & Resources, 2021, 33(1): 45-53.
链接本文:  
https://www.gtzyyg.com/CN/10.6046/gtzyyg.2020162      或      https://www.gtzyyg.com/CN/Y2021/V33/I1/45
Fig.1  ResNet基本结构
Fig.2  Deeplab v3语义分割网络结构[17]
Fig.3  PSPNet语义分割网络结构[20]
Fig.4  ShelfNet语义分割网络结构[21]
Fig.5  部分训练数据集
Fig.5-2  部分训练数据集
网络 训练集平
均损失
验证集平
均损失
验证集平
均准确率
mIOU 背景IOU 前景IOU 前20循环平
均前景IOU
Deeplab v3-50 19.91 12.15 87.90 82.21 94.30 70.13 60.35
Deeplab v3-101 18.92 12.37 87.00 81.65 94.19 69.11 58.05
PSPNet50 19.21 12.02 88.03 82.35 94.34 70.36 63.30
PSPNet101 18.33 12.22 87.92 82.27 94.32 70.23 63.27
ShelfNet50 18.01 12.48 89.19 82.75 94.33 71.16 63.60
ShelfNet101 21.17 13.15 86.82 81.34 94.08 68.60 58.23
Tab.1  网络前期训练精度比较
分类器/网络 训练集平
均损失/%
验证集平
均损失/%
验证集平
均准确率/%
mIOU/% 背景IOU/% 前景IOU/% 51-80循环平
均前景IOU/%
训练时
长/h
随机森林 85.48 55.98
支持向量机 79.82 46.27
Deeplab v3-50 13.85 10.13 90.65 85.38 95.30 75.46 74.74 48.92
Deeplab v3-101 12.78 9.68 90.88 85.70 95.41 75.98 75.68 68.35
PSPNet50 13.72 9.81 90.78 85.37 95.28 75.45 74.73 32.13
PSPNet101 10.95 8.66 92.44 87.25 95.88 78.63 78.01 51.86
ShelfNet50 11.54 9.08 92.38 86.71 95.65 77.76 77.05 14.02
ShelfNet101 12.05 9.31 91.60 86.12 95.49 76.94 76.89 15.63
Tab.2  网络最终训练精度比较
Fig.6-1  ShelfNet50网络测试集分割结果
Fig.6-2  ShelfNet50网络测试集分割结果
Fig.7  ShelfNet50网络测试集分割误差
[1] 中华人民共和国建设部. 城市规划基本术语标准:GB/T50280—1998[S]. 北京:工程建设标准全文信息系统, 1999.
Ministry of Construction of the People’s Republic of China. Basic terminology standard for urban planning:GB/T50280—1998[S]. Beijing:Engineering Construction Standard Full-Text Information System, 1999.
[2] Sinha P, Verma N K, Ayele E. Urban built-up area extraction and change detection of Adama municipal area using time-series Landsat images[J]. International Journal of Advanced Remote Sensing and GIS, 2016,5(8):1886-1895.
[3] Zha Y, Gao J, Ni S. Use of normalized difference built-up index in automatically mapping urban areas from TM imagery[J]. International Journal of Remote Sensing, 2003,24(3):583-594.
[4] 陈志强, 陈健飞. 基于NDBI指数法的城镇用地影像识别分析与制图[J]. 地球信息科学学报, 2012,8(2):137-140.
Chen Z Q, Chen J F. Urban land image recognition analysis and mapping based on NDBI index method[J]. Geo-Information Science, 2012,8(2):137-140.
[5] 梁珊, 刘荣, 祖帅, 等. 基于多光谱遥感影像的城市建设用地提取及扩张分析[J]. 测绘与空间地理信息, 2018,41(6):74-78.
Liang S, Liu R, Zu S, et al. Urban construction land extraction and expansion analysis based on multi-spectral remote sensing images[J]. Geomatics and Spatial Geographic Information, 2018,41(6):74-78.
[6] Liaw A, Wiener M. Classification and regression by random forest[J]. R News, 2002,2(3):18-22.
[7] Burges C J C. A tutorial on support vector machines for pattern reco-gnition[J]. Data Mining and Knowledge Discovery, 1998,2(2):121-167.
[8] Gowthami K, Thilagavathi K. Built-up area detection of remote sensing images using static clustering technique[C]// 2014 International Conference on Electronics and Communication Systems(ICECS).IEEE, 2014: 1-5.
[9] Gong P, Liu H, Zhang M, et al. STab.classification with limited sample:Transferring a 30-m resolution sample set collected in 2015 to mapping 10-m resolution global land cover in 2017[J]. Science Bulletin, 2019,64(6):370-373.
doi: 10.1016/j.scib.2019.03.002
[10] LeCun Y, Boser B, Denker J S, et al. Backpropagation applied to handwritten zip code recognition[J]. Neural Computation, 1989,1(4):541-551.
[11] Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks[C]// Advances in Neural Information Processing Systems, 2012: 1097-1105.
[12] 杨建宇, 周振旭, 杜贞容, 等. 基于SegNet语义模型的高分辨率遥感影像农村建设用地提取[J]. 农业工程学报, 2019,35(5):259-266.
Yang J Y, Zhou Z X, Du Z R, et al. High-resolution remote sensing imagery rural construction land extraction based on SegNet semantic model[J]. Transactions of the Chinese Society of Agricultural Engineering, 2019,35(5):259-266.
[13] Badrinarayanan V, Kendall A, Cipolla R. Segnet:A deep convolutional encoder-decoder architecture for image segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017,39(12):2481-2495.
doi: 10.1109/TPAMI.2016.2644615 pmid: 28060704
[14] 苏健民, 杨岚心, 景维鹏. 基于U-Net的高分辨率遥感图像语义分割方法[J]. 计算机工程与应用, 2019,55(7):212-218.
Su J M, Yang L X, Jing W P. High-resolution remote sensing image semantic segmentation method based on U-Net[J]. Computer Engineering and Applications, 2019,55(7):212-218.
[15] Ronneberger O, Fischer P, Brox T. U-net:Convolutional networks for biomedical image segmentation[C]// International Conference on Medical Image Computing and Computer-Assisted Intervention.Springer,Cham.Berlin,German:Springer, 2015: 234-241.
[16] Chen L C, Papandreou G, Schroff F. et al. Rethinking atrous convolution for semantic image segmentation[EB/OL].12017-12-05. http://arxiv.org/odf/1706.05587.pdf.
[17] Yu F, Koltun V. Multi-scale context aggregation by dilated convolutions[J/OL]. [2019-05-30].https://arxiv.org/abs/1511.07122.pdf.
[18] 王俊强, 李建胜, 周华春, 等. 基于Deeplabv3+与CRF的遥感影像典型要素提取方法[J]. 计算机工程, 2019,45(10):260-265,271.
Wang J Q, Li J S, Zhou H C, et al. Typical elements extraction method of remote sensing image based on Deeplabv3+ and CRF[J]. Computer Engineering, 2019,45(10):260-265,271.
[19] He K M, Zhang X Y, Ren S Q, et al. Deep residual learning for image recognition[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE, 2016: 770-778.
[20] Zhao H S, Shi J P, Qi X J, et al. Pyramid scene parsing network[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE, 2017: 2881-2890.
[21] Zhuang J T, Yang J L, Gu L, et al. ShelfNet for real-time semantic segmentation[EB/OL].(2018-11-27). http://arxiv.org/pdf/1811.11254 vl.pdf.
[22] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[EB/OL].(2014-09-04). http://arxiv.org/pdf//1409.1556 vl.pdf.
[1] 刘立, 董先敏, 刘娟. 顾及地学特征的遥感影像语义分割模型性能评价方法[J]. 自然资源遥感, 2023, 35(3): 80-87.
[2] 林佳惠, 刘广, 范景辉, 赵红丽, 白世彪, 潘宏宇. 联合改进U-Net模型和D-InSAR技术采矿沉陷提取方法[J]. 自然资源遥感, 2023, 35(3): 145-152.
[3] 郑宗生, 刘海霞, 王振华, 卢鹏, 沈绪坤, 唐鹏飞. 改进3D-CNN的高光谱图像地物分类方法[J]. 自然资源遥感, 2023, 35(2): 105-111.
[4] 胡建文, 汪泽平, 胡佩. 基于深度学习的空谱遥感图像融合综述[J]. 自然资源遥感, 2023, 35(1): 1-14.
[5] 赵凌虎, 袁希平, 甘淑, 胡琳, 丘鸣语. 改进Deeplabv3+的高分辨率遥感影像道路提取模型[J]. 自然资源遥感, 2023, 35(1): 107-114.
[6] 孙盛, 蒙芝敏, 胡忠文, 余旭. 多尺度轻量化CNN在SAR图像地物分类中的应用[J]. 自然资源遥感, 2023, 35(1): 27-34.
[7] 孟琮棠, 赵银娣, 韩文泉, 何晨阳, 陈锡秋. 基于RandLA-Net的机载激光雷达点云城市建筑物变化检测[J]. 自然资源遥感, 2022, 34(4): 113-121.
[8] 沈骏翱, 马梦婷, 宋致远, 柳汀洲, 张微. 基于深度学习语义分割模型的高分辨率遥感图像水体提取[J]. 自然资源遥感, 2022, 34(4): 129-135.
[9] 杨昭颖, 韩灵怡, 郑向向, 李文吉, 冯磊, 王轶, 杨永鹏. 基于卷积神经网络的遥感影像及DEM滑坡识别——以黄土滑坡为例[J]. 自然资源遥感, 2022, 34(2): 224-230.
[10] 孔爱玲, 张承明, 李峰, 韩颖娟, 孙焕英, 杜漫飞. 基于知识引导的遥感影像融合方法[J]. 自然资源遥感, 2022, 34(2): 47-55.
[11] 王华俊, 葛小三. 一种轻量级的DeepLabv3+遥感影像建筑物提取方法[J]. 自然资源遥感, 2022, 34(2): 128-135.
[12] 廖廓, 聂磊, 杨泽宇, 张红艳, 王艳杰, 彭继达, 党皓飞, 冷伟. 基于多维卷积神经网络的多源高分辨率卫星影像茶园分类[J]. 自然资源遥感, 2022, 34(2): 152-161.
[13] 于新莉, 宋妍, 杨淼, 黄磊, 张艳杰. 结合空间约束的卷积神经网络多模型多尺度船企场景识别[J]. 自然资源遥感, 2021, 33(4): 72-81.
[14] 刘万军, 高健康, 曲海成, 姜文涛. 多尺度特征增强的遥感图像舰船目标检测[J]. 自然资源遥感, 2021, 33(3): 97-106.
[15] 郭文, 张荞. 基于注意力增强全卷积神经网络的高分卫星影像建筑物提取[J]. 国土资源遥感, 2021, 33(2): 100-107.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
京ICP备05055290号-2
版权所有 © 2015 《自然资源遥感》编辑部
地址:北京学院路31号中国国土资源航空物探遥感中心 邮编:100083
电话:010-62060291/62060292 E-mail:zrzyyg@163.com
本系统由北京玛格泰克科技发展有限公司设计开发