Please wait a minute...
img

官方微信

遥感技术与应用  2022, Vol. 37 Issue (4): 789-799    DOI: 10.11873/j.issn.1004-0323.2022.4.0789
深度学习专栏     
基于深度学习的高分辨率卫星遥感影像围填海检测识别
于枫世1(),隋毅1(),王常颖1,初佳兰2
1.青岛大学计算机科学技术学院,山东 青岛 266071
2.国家海洋环境监测中心,辽宁 大连 116023
Reclamation Detection and Recognition of High Resolution Satellite Remote Sensing Image based on Deep Learning
Fengshi Yu1(),Yi Sui1(),Changying Wang1,Jialan Chu2
1.School of Computer Science and Technology,Qingdao University,Qingdao 266071,China
2.School of Data Science and Software Engineering,Qingdao University,Qingdao 266071,China
 全文: PDF(6175 KB)   HTML
摘要:

基于高分辨率卫星遥感影像自动、准确提取围填海土地利用现状,是实现围填海集约使用的重要技术手段。针对高分辨率卫星遥感影像地物特征复杂,依赖人工提取特征的传统方法较难满足业务部门实际需求的问题,提出了基于深度学习的围填海检测识别技术框架,该框架使用U-Net网络的多约束变体结构,并针对高分辨率遥感影像地物特征复杂导致地物分类不一致的问题,引入全连接条件随机场和图像腐蚀运算对分割结果进行后处理。以天津市滨海新区2016年和2020年高分辨卫星遥感影像为数据源进行了验证,实验表明围填海地物分割整体准确率、F1-score、Kappa系数以及mIoU分别达到96.73%、92.87%、90.28%、86.82%。在此基础上,分析提取了该围填海区域土地利用动态变化特征,为围填海集约使用管理提供了有效技术支撑。

关键词: 围填海深度学习检测识别U?Net    
Abstract:

Based on high-resolution satellite remote sensing images, automatic and accurate extraction of land use status of reclamation is an important technical means to realize the intensive use of reclamation. In view of the complex features of high-resolution satellite remote sensing images, the traditional method of manually extracting features is difficult to meet the actual needs of business departments. A framework of reclamation detection and recognition based on deep learning is proposed. The framework uses the multi constrained variant structure of U-Net network, and to solve the problem of inconsistent classification caused by complex features of high-resolution remote sensing images, full connection conditional random field and image corrosion operation are introduced to post-processing the segmentation results. The high-resolution satellite remote sensing images of Tianjin Binhai New Area in 2016 and 2020 were used as data sources to verify. The experimental results show that the overall accuracy rate, F1 score, kappa coefficient and mIoU of reclamation are 96.73%, 92.87%, 90.28% and 86.82% respectively. On this basis, the dynamic change characteristics of land use in the reclamation area are analyzed and extracted, which provides effective technical support for the intensive use and management of reclamation.

Key words: Reclamation    Deep learning    Detection and identification    U-Net
收稿日期: 2020-08-31 出版日期: 2022-09-28
:  P756.8  
基金资助: 国家自然科学青年基金项目(41706198);国家自然科学基金面上项目(41876109);山东省高等学校科技计划项目(J17KA056)
通讯作者: 隋毅     E-mail: 2019020589@qdu.edu.cn;suiyi@qdu.edu.cn
作者简介: 于枫世(1997-),男,山东烟台人,硕士研究生,主要从事遥感图像处理。E?mail:2019020589@qdu.edu.cn
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  
于枫世
隋毅
王常颖
初佳兰

引用本文:

于枫世,隋毅,王常颖,初佳兰. 基于深度学习的高分辨率卫星遥感影像围填海检测识别[J]. 遥感技术与应用, 2022, 37(4): 789-799.

Fengshi Yu,Yi Sui,Changying Wang,Jialan Chu. Reclamation Detection and Recognition of High Resolution Satellite Remote Sensing Image based on Deep Learning. Remote Sensing Technology and Application, 2022, 37(4): 789-799.

链接本文:

http://www.rsta.ac.cn/CN/10.11873/j.issn.1004-0323.2022.4.0789        http://www.rsta.ac.cn/CN/Y2022/V37/I4/789

图1  经典U-Net网络结构[16]
图2  多约束变体U-Net(MCFCN)网络结构
图3  全连接CRF处理流程
图4  后处理结果对比
图5  围填海检测流程图
图6  研究区域和测试集、训练集区域的划分审图号:GS(2019)3266
模型αβγδ
MCFCN10.50.500
MCFCN20.500.50
MCFCN30.5000.5
MCFCN40.250.250.250.25
MCFCN51000
表1  损失函数中约束参数选取
模型总体分类精度F1-scoreKappamIoU
MCFCN10.9720.9440.9240.895
MCFCN20.9730.9450.9260.897
MCFCN30.9720.9420.9230.892
MCFCN40.9730.9430.9240.893
MCFCN50.9710.9400.9200.889
表2  不同参数条件下验证集上模型性能比较
时间模型总体分类精度F1-scoreKappamIoU
2016年SVM0.913 60.723 60.657 90.605 0
FCN8s0.928 40.839 70.779 80.729 2
U-Net0.957 40.908 60.872 80.834 4
Linknet0.957 70.910 00.874 30.836 7
MCFCN20.958 10.915 30.879 70.844 8
MCFCN2+后处理0.962 10.924 30.892 00.860 1

2020年

SVM0.876 00.684 90.589 00.547 5
FCN8s0.931 70.844 00.789 90.735 8
U-Net0.955 50.905 50.870 40.829 1
Linknet0.955 70.904 00.869 40.826 9
MCFCN20.961 20.914 10.883 50.843 7
MCFCN2+后处理0.967 30.928 70.902 80.868 2
表3  测试区域结果比较
模型参数量(106训练速度(FPS)
UNet3.408 994.77
Linknet8.328 675.32
MCFCN23.414 888.31
表4  模型参数规模
图7  2016年和2020年测试区域结果比较
图8  添加后处理结果比较
图9  2016年和2020年研究区域划分及结果 审图号:GS(2019)3266
2016年2020年
已利用/km2未利用/km2已利用比例已利用/km2未利用/km2已利用比例
滨海旅游区0.46422.3762.03%1.55721.1036.87%
东疆港区10.23925.68428.50%10.79224.68330.42%
北疆港区25.4764.70284.42%24.5865.31582.22%
南疆港区15.42111.00258.36%15.34911.33657.52%
临港工业区115.67056.71221.65%17.84057.22523.77%
临港工业区20.42028.6961.44%0.29722.5101.30%
整体67.690149.17231.21%70.421142.17233.12%
表5  2016—2020年各港区土地利用情况
图10  2016年和2020年各港区陆地面积和利用土地面积
1 Suo Anning, Wang Peng, Yuan Daowei,et al. Study on monitoring and analysis of existing sea reclamation resource based on high resolution satellite remote sensing imagery:A case in south coast of Yingkou[J]. Acta Oceanologica Sinica,2016,38(9):54-63.
1 索安宁, 王鹏,袁道伟,等. 基于高空间分辨率卫星遥感影像的围填海存量资源监测与评估研究——以营口市南部海岸为例[J]. 海洋学报,2016,38(9):54-63.]
2 Gao Ning, Du Xinyuan, Xu Peng,et al. Monitoring technology and method of doubtful area in reclamation area[J]. Ocean Development and Management,2019,36(2):14-16.
2 高宁, 杜新远,许鹏,等. 围填海疑点疑区监测技术方法[J]. 海洋开发与管理,2019,36(2):14-16.]
3 Sun Jing, Zhang Ying, Qiao Qinghua,et al. Reclamation project monitoring and ecosystem service value change analysis[J]. Science of Surveying and Mapping,2020,45(12):201-208.
3 孙静,张英,乔庆华,等. 围填海工程监测及生态系统服务价值变化分析[J].测绘科学,2020,45(12):201-208.]
4 Jia Wei, Gao Xiaohong, Yang Lingyu,et al. Land cover information extraction for complicated terrain regions via an object-oriented classification method[J]. Journal of Lanzhou University(Natural Sciences Edition),2018,54(4):486-493.
4 贾伟,高小红,杨灵玉,等. 面向对象方法的复杂地形区地表覆盖信息提取[J].兰州大学学报(自然科学版),2018,54(4):486-493.]
5 Ju Mingming, Wang Min, Zhang Dong,et al. Study on the remote-sensing-based monitoring of reclamation project area by using the object-oriented image analysis technique[J]. Marine Science Bulletin,2013,32(6):678-684.
5 鞠明明,汪闽,张东,等. 基于面向对象图像分析技术的围填海用海工程遥感监测[J]. 海洋通报,2013,32(6):678-684.]
6 Zhou Tianya.Object-oriented land cover classification using high spatial resolution remote sensing[J]. Geomatics Science and Technology,2020,8(1):9-16.
6 周天涯. 面向对象的高空间分辨率遥感土地覆盖分类[J].测绘科学技术,2020,8(1): 9-16.]
7 Rao Ping, Wang Jianli, Wang Yong,et al. Extraction of information on construction land based on multi-feature decision tree classificationExtraction of information on construction land based on multi-feature decision tree classification[J]. Transactions of the Chinese Society of Agricultural Engineering,2014,30(12):233-240.
7 饶萍,王建力,王勇. 基于多特征决策树的建设用地信息提取[J]. 农业工程学报,2014,30(12):233-240.]
8 Zhang Hongmei, Wu Jiwen, Liu Xing,et al. Land use classification based on feature extraction and decision tree method[J]. Science of Surveying and Mapping, 2014,39(10):53-56.
8 张红梅,吴基文,刘星,等. 特征提取和决策树法土地利用遥感分类[J]. 测绘科学,2014,39(10):53-56.]
9 Zhang Qian. Research on land cover classification of aerial hyperspectral remote sensing based on decision tree method[D]. Ji'nan: Shandong University of Science and Technology,2015[张倩. 基于决策树方法的航空高光谱遥感土地覆盖分类研究[D]. 济南:山东科技大学,2015.]
10 Shen Zhaoqing, Huang Liang, Tao Jianbin. Hyperspectral RS image road feature extraction based on SVM[J]. Journal of Chang’an University (Natural Science Edition),2012,32(5): 34-38.
10 沈照庆,黄亮,陶建斌. 基于支持向量机的高光谱遥感影像道路提取[J]. 长安大学学报(自然科学版),2012,32(5):34-38.]
11 Yuan Yongsheng. Research on object-oriented land cover image classification method based on support vector machine classification[D].Xianyang:Northwest A&F University.
11 员永生. 基于支持向量机分类的面向对象土地覆被图像分类方法研究[D]. 咸阳:西北农林科技大学,2010.
12 Zhu Jieer. Research on support vector machine classification of hyperspectral images combined with spatial information[D]. Hangzhou: Zhejiang University.
12 朱洁尔. 结合空间信息的高光谱图像支持向量机分类研究[D]. 杭州:浙江大学,2013.
13 Le Cun, Bengio Yoshua. Convolutional networks for images, speech, and time series[J]. The Handbook of Brain Theory and Neural Networks, 1995,3361(10):255-258.
14 Cao Linlin, Li Haitao,HanYanshun, et al. Application of convolutional neural networks in classification of high resolution remote sensing imagery[J]. Science of Surveying and Mapping, 2016,41(9):170-175.
14 曹林林,李海涛,韩颜顺,等. 卷积神经网络在高分遥感影像分类中的应用[J]. 测绘科学,2016,41(9):170-175.
15 Long J, Shelhamer E, Darrell T. Fully convolutional networks for semantic segmentation[C]∥ 2015 IEEE Conference on Computer Vision and Pattern Recognition(CVPR), 2015: 3431-3440.DOI: 10.1109/CVPR.2015.7298965 .
doi: 10.1109/CVPR.2015.7298965
16 Ronneberger O, Fischer O, Brox T. U-Net: convolutional networks for biomedical image segmentation[J]. Medical Image Computing and Computer-Assisted Intervention, Pt Iii, 2015,9351:234-241. DOI:10.1007/978-3-319-24574-4_28 .
doi: 10.1007/978-3-319-24574-4_28
17 Su Jianmin, Yang Lanxin, Jing Weipeng. U-Net based semantic segmentation method for high resolution remote sensing image[J]. Computer Engineering and Applications, 2019,55(7):207-213.
17 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):207-213.]
18 Xu Huimin. Research on high resolution remote sensing image classification method based on deep learning U-Net model[D]. Chengdu: Southwest Jiaotong University.
18 许慧敏. 基于深度学习U-Net模型的高分辨率遥感影像分类方法研究[D]. 成都:西南交通大学,2018.
19 Chaurasia A, Culurciello E. LinkNet: Exploiting encoder representations for dfficient semantic segmentation[C]∥2017 IEEE Visual Communications and Image Processing (VCIP),2017. DOI: 10.1109/ VCIP.2017.8305148 .
doi: 10.1109/ VCIP.2017.8305148
20 Li L, Wang C, Zhang H, et al. Residual unet for urban building change detection with Sentinel-1 SAR data[J]. 2019 IEEE International Geoscience and Remote Sensing Symposium(IGARSS 2019),2019:1498-1501. DOI:10.1109/IGARSS. 2019.8898146 .
doi: 10.1109/IGARSS. 2019.8898146
21 Sergey I, Christian S. Batch normalization: Accelerating deep network training by reducing internal covariate shift[C]∥ International Conference on Machine Learning. PMLR, 2015: 448-456. DOI: 10.5555/3045118.3045167 .
doi: 10.5555/3045118.3045167
22 Wu G M, Shao X W, Guo Z L, et al. Automatic building segmentation of aerial imagery using multi-constraint fully convolutional networks[J]. Remote Sensing,2018,10:407. DOI:10. 3390/rs10030407 .
doi: 10. 3390/rs10030407
23 Chen L C, Papandreou G, Kokkinos I,et al. DeepLab: semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected CRFs[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2018,40(4):834-848. DOI:10.1109/TPAMI.2017.2699184 .
doi: 10.1109/TPAMI.2017.2699184
24 Bittner K, Cui S, Reinartz P. Building extraction from remote sensing data using fully convolutional networks[J]. ISPRS Hannover Workshop: Hrigi 17-Cmrt 17-Isa 17-Eurocow 17, 2017,42-1(W1):481-486. DOI: 10.5194/isprs-archives-XLII-1-W1-481-2017 .
doi: 10.5194/isprs-archives-XLII-1-W1-481-2017
25 Mark Everingham, Gool Luc Van, Williams Christopher K I, et al. The pascal Visual Object Classes (VOC) challenge[J]. 2010,8(2):303-338. DOI: 10.1007/s11263-009-0275-4 .
doi: 10.1007/s11263-009-0275-4
26 Polak M, Zhang H, Pi M H. An evaluation metric for image segmentation of multiple objects[J]. Image and Vision Computing, 2009,27(8):1223-1227. DOI:10.1016/j.imavis. 2008.09.008 .
doi: 10.1016/j.imavis. 2008.09.008
27 Carletta J. Assessing agreement on classification tasks: The Kappa statistic[J].Computational Linguistics,1996,22(2): 249-254. DOI:10.5555/230386.230390 .
doi: 10.5555/230386.230390
[1] 史佳睿,申茜,彭红春,李利伟,姚月,汪明秀,王茹. 基于三种高分辨率影像样本迁移学习的城市水体提取研究[J]. 遥感技术与应用, 2022, 37(3): 731-738.
[2] 董秀春,刘忠友,蒋怡,郭涛,李宗南. 基于WorldView-2影像和语义分割模型的小麦分类提取[J]. 遥感技术与应用, 2022, 37(3): 564-570.
[3] 付涵,范湘涛,严珍珍,杜小平. 基于深度学习的遥感图像目标检测技术研究进展[J]. 遥感技术与应用, 2022, 37(2): 290-305.
[4] 吕冬梅,马玥,李华朋. 基于CNN的吉林一号卫星城市土地覆被制图潜力评估[J]. 遥感技术与应用, 2022, 37(2): 368-378.
[5] 林广坤,吴志峰,曹峥,关文川. 基于SBAS-InSAR技术的围填海区域地面沉降监测[J]. 遥感技术与应用, 2021, 36(6): 1358-1367.
[6] 路春燕,雷依凡,苏颖,黄雨菲,刘明月,贾明明. 基于面向对象—深度学习的闽东南低海拔海岸带地区湿地动态遥感分析[J]. 遥感技术与应用, 2021, 36(4): 713-727.
[7] 郑磊,何直蒙,丁海勇. 基于ENVINet5的高分辨率遥感影像稀疏塑料大棚提取研究[J]. 遥感技术与应用, 2021, 36(4): 908-915.
[8] 李庆,陈俊杰,李庆亭,李柏鹏,卢凯旋,昝露洋,陈正超. 基于SSD模型的京津冀地区尾矿库检测[J]. 遥感技术与应用, 2021, 36(2): 293-303.
[9] 林娜,冯丽蓉,张小青. 基于优化Faster-RCNN的遥感影像飞机检测[J]. 遥感技术与应用, 2021, 36(2): 275-284.
[10] 陈妮,应丰,王静,李健. 基于U-Net的高分辨率遥感图像土地利用信息提取[J]. 遥感技术与应用, 2021, 36(2): 285-292.
[11] 胡乃勋,陈涛,甄娜,牛瑞卿. 基于卷积神经网络的面向对象露天采场提取[J]. 遥感技术与应用, 2021, 36(2): 265-274.
[12] 张永宏,曹海啸,阚希. 基于FY-4A/AGRI时空特征融合的新疆地区积雪判识[J]. 遥感技术与应用, 2020, 35(6): 1337-1347.
[13] 徐瑾昊,冯敏,王建邦,冉有华,祁元,杨联安,李新. 基于高分遥感数据和深度学习的石冰川自动提取研究[J]. 遥感技术与应用, 2020, 35(6): 1329-1336.
[14] 杨瑞,祁元,苏阳. 深度学习U-Net方法及其在高分辨卫星影像分类中的应用[J]. 遥感技术与应用, 2020, 35(4): 767-774.
[15] 张峰极,吴艳兰,姚雪东,梁泽毓. 基于改进DenseNet网络的多源遥感影像露天开采区智能提取方法[J]. 遥感技术与应用, 2020, 35(3): 673-684.