Please wait a minute...
img

官方微信

遥感技术与应用  2022, Vol. 37 Issue (4): 811-819    DOI: 10.11873/j.issn.1004-0323.2022.4.0811
深度学习专栏     
基于多尺度特征融合的U-Net网络高分影像不透水面提取研究
王晶1(),高帅2(),郭亮3,4,汪云5
1.中国地质大学 地球科学与资源学院,北京 100083
2.中国科学院空天信息创新研究院,北京 100094
3.广州市城市规划勘测设计研究院,广东 广州 510060
4.广东省城市感知与监测预警企业重点实验室,广东 广州 510060
5.北京林业大学 园林学院,北京 100083
Impervious Surface Extraction from High-resolution Images based on Multi-scale Feature Fusion in U-Net Network
Jing Wang1(),Shuai Gao2(),Liang Guo3,4,Yun Wang5
1.School of Earth Sciences and Resources,China University of Geosciences,Beijing 100083,China
2.Aerospace Information Research Institute,Chinese Academy of Sciences,Beijing 100094,China
3.Guangzhou Urban Planning & Design Survey Research Institute,Guangzhou 510060,China
4.Guangdong Enterprise Key Laboratory for Urban Sensing,Monitoring and Early Warning,Guangzhou 510060,China
5.The College of Forestry of Beijing Forestry University,Beijing 100083,China
 全文: PDF(2535 KB)   HTML
摘要:

不透水面作为监测城市生态环境的重要指标,其信息提取具有重要意义。由于城市地表的复杂性及细化的城市管理需要,急需提取高精度的城市不透水面。但是基于传统方法提取高精度的城市不透水面面临巨大困难。而深度学习方法因其自动化提取影像特征的特点逐渐成为遥感影像地物提取的新兴方法。基于此,采用多尺度特征融合的U-Net深度学习方法以提升语义分割精度,开展高分辨率遥感影像不透水面的精确提取研究。模型引入残差模块代替普通卷积以加深网络,提取更多影像特征;加入金字塔池化模块增强网络对复杂场景的解析能力;利用跳跃连接方式融合不同尺度特征,有利于恢复空间信息。以广州市航摄正射影像为数据源,通过卷积神经网络将遥感影像分割为背景、其他、植被、道路和房屋5种地物类型,将其与人工目视解译的地面真值进行验证,最终提取研究区域不透水面。实验证明:多尺度特征融合的U-Net模型总体精度和Kappa系数分别为87.596% 和0.82。在定性与定量两个方面均优于传统的监督分类法、面向对象分类法和经典U-Net模型法。结果表明:该模型利用多维度影像特征信息,有效提升了复杂场景图像的分割精度,分割效果好,适用于高分辨率遥感影像不透水面提取,该研究成果可为城市环境监测提供数据支撑。

关键词: 不透水面U-Net模型残差模块金字塔池化模块高分辨率遥感    
Abstract:

As an important index for monitoring urban ecological environment, the extraction of impervious surface is of great significance. Due to the complexity of urban surface and the need of detailed urban management, it is urgent to extract high-precision urban impervious surface. However, it is difficult to extract high precision impervious surface based on traditional methods. Deep learning method has gradually become a new method of remote sensing image feature extraction because of its characteristics of automatic image feature extraction. Based on this, this paper uses the U-Net deep learning method based on multi-scale feature fusion to improve the semantic segmentation accuracy, and carries out the research on the accurate extraction of impervious surface from high resolution remote sensing images.The residual module is introduced instead of convolution to deepen the network and extract more image features, the pyramid pooling module is added to enhance the network's ability to resolve complex scenarios. It is beneficial to recover spatial information by combining different scale features with jump connection. In this paper, aerial orthophoto images of Guangzhou were taken as the data source. Through convolutional neural network, the remote sensing image is segmtioned into five types of features: background, others, vegetation, road and building. Verify it with the ground truth value of manual visual interpretation, finally, the impervious surface of the study area was extracted. Experiments show that the overall accuracy and Kappa coefficient of the U-NET model are 87.596% and 0.82, respectively. It is superior to traditional supervised taxonomy, object-oriented taxonomy and classical U-Net model in both qualitative and quantitative aspects. The results show that the model can effectively improve the segmentation accuracy of complex scene images by using the multi-dimensional image feature information, and the segmentation effect is good, which is suitable for the extraction of impervious water from high resolution remote sensing images. The research results in this paper can provide data support for urban environmental monitoring.

Key words: Impervious surface    U-Net    Residual module    Pyramid pooling module    High resolution remote sensing
收稿日期: 2021-09-08 出版日期: 2022-09-28
:  TP79  
基金资助: 国家自然科学基金项目(42171377);国家重点研发计划(2017YFA0603004);高分项目(30-Y20A15-9003-17/18);广东省重点领域研发计划(2020B0101130009);广东省城市感知与监测预警企业重点实验室基金项目(2020B121202019)
通讯作者: 高帅     E-mail: wj970205@126.com;gaoshuai@radi.ac.cn
作者简介: 王 晶(1997-),女,天津人,硕士研究生,主要从事机器学习与遥感信息处理研究。E?mail:wj970205@126.com
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  
王晶
高帅
郭亮
汪云

引用本文:

王晶,高帅,郭亮,汪云. 基于多尺度特征融合的U-Net网络高分影像不透水面提取研究[J]. 遥感技术与应用, 2022, 37(4): 811-819.

Jing Wang,Shuai Gao,Liang Guo,Yun Wang. Impervious Surface Extraction from High-resolution Images based on Multi-scale Feature Fusion in U-Net Network. Remote Sensing Technology and Application, 2022, 37(4): 811-819.

链接本文:

http://www.rsta.ac.cn/CN/10.11873/j.issn.1004-0323.2022.4.0811        http://www.rsta.ac.cn/CN/Y2022/V37/I4/811

图1  技术路线图
图2  数据集影像及其标签
图3  残差模块
图4  多尺度特征融合U-Net网络
项目硬件系统CPUGPUPytorchPythonCUDA
内容Ubuntu 18.04Intel E5-2603NVIDIA TITAN X (Pascal)1.63.610.1
表1  实验环境配置
评价指标计算方法
IoU

IoU =?|A?B||A?B|

其中A代表预测结果的像素集合,B代表真实值的像素集合。

Kappa系数

Kappa = p0-pe1-pe

Pe =?a1*b1+a2*b2++an*bnN*N

其中P0表示每一类正确分类的像元数占总像元的比例,a代表每个类别的实际像元数,b代表每个类别的预测像元数,N代表总像元数。

总体分类精度正确分类的像元数占总像元数的比例
准确率被正确分类的像元数占预测为正确的总像元数的比例
召回率被正确分类的像元数占实际正确的总像元数的比例
F1分数F1=2*precision*recallprecision+recall
表2  分类结果评价指标
精度召回率F1分数IoUKappa总体精度
背景0.940.650.770.620.8287.596%
其他0.180.160.170.09
植被0.910.940.920.85
道路0.740.860.800.66
房屋0.920.880.900.81
表3  模型精度评价
图5  越秀区不透水面提取结果
原图最大似然法面向对象法经典U-net法本文方法
区域1
区域2
区域3
区域4
区域5
表4  研究区提取结果比较
1 Arnold C L, Gibbons C J. Impervious surface coverage: The emergence of a key environmental indicator[J]. Journal of the American Planning Association, 1996,62(2):243-258.
2 Weng Q. Remote sensing of impervious surfaces in the urban areas: Requirements, methods, and trends[J]. Remote Sensing of Environment, 2012,117:34-49. DOI:10.1016/j.rse. 2011.02.030 .
doi: 10.1016/j.rse. 2011.02.030
3 Peng Wenfu, Zhang Donghui, He Zhengwei,et al. Study on response of urban surface temperature to impervious surface area in the city Chengdu[J].Remote Sensing Information,2010,25(2):98-102.
3 彭文甫,张东辉,何政伟,等.成都市地表温度对不透水面的响应研究[J].遥感信息,2010,25(2):98-102.
4 Arnold C L, Gibbons C J. Impervious surface coverage: The emergence of a key environmental indicator[J]. Journal of the American Planning Association, 1996,62(2):243-258.
5 Zhang H S, Lin H, Wang Y P. A new scheme for urban impervious surface classification from SAR images[J]. ISPRS Journal of Photogrammetry and Remote Sensing,2018(139):103-118.
6 Deng C B, Wu C S. A spatially adaptive spectral mixture analysis for mapping subpixel urban impervious surface distribution[J].Remote Sensing of Environment,133:62-70. DOI: 10. 1016/j.rse.2013.02.005 ]
doi: 10. 1016/j.rse.2013.02.005
7 Xu Hanqiu, Wang Meiya.Remote sensing-based retrieval of ground impervious surfaces[J]. Journal of Remote Sensing,2016,20(5):1270-1289
7 徐涵秋,王美雅.地表不透水面信息遥感的主要方法分析[J].遥感学报,2016,20(5):1270-1289.
8 Cao Yong, Tao Yuxiang, Deng Lu, et al. An impervious surface index construction for restraining bare land[J]. Remote Sensing for Land & Resourses, 2020,32(3):71-79.
8 曹勇,陶于祥,邓陆,等.一种抑制裸地的不透水面指数构建[J].国土资源遥感,2020,32(3):71-79.
9 Yang Xu, Zhu Daming, Yang Runshu, et al.A visible-band remote sensing index for extracting impervious surfaces[J].Transactions of the Chinese Society of Agricultural Engineering, 2020,36(8):127-134.
9 杨栩,朱大明,杨润书,等.一种提取不透水面的可见光波段遥感指数[J].农业工程学报,2020,36(8):127-134.
10 Ridd M K. Exploring a V-I-S (Vegetation-Impervious Surface Soil) model for urban ecosystem analysis through remote sensing: comparative anatomy for cities[J]. International Journal of Remote Sensing,1995, 16(12): 2165-2185.
11 Wu C S, Murrary A T. Estimating impervious surface distribution by spectral mixture analysis[J]. Remote Sensing of Environment,84(4):493-505. DOI:10.1016/S0034-4257(02)00136-0 .
doi: 10.1016/S0034-4257(02)00136-0
12 Li Fanggang, Li Erzhu,Samat Alim et al. Estimation of large-scale impervious surface percentage by fusion of multi-source time series remote sensing data[J]. Journal of Remote Sensing, 2020,24(10):1243-1254.
12 李方刚,李二珠,阿里木·赛买提,等.融合多源时序遥感数据大尺度不透水面覆盖率估算[J].遥感学报,2020,24(10):1243-1254.
13 He Guizhen, Cheng Xiaojun. Study on monitoring and evaluating of flood disaster based on RS and GIS[J]. Remote Sensing Information,2012,27(1): 84-87.
13 何桂珍, 程效军. 基于RS 和GIS 的洪涝灾害监测与评估研究[J]. 遥感信息,2012,27(1):84-87.
14 Zuo Jiaqi, Wang Zegen, Bian Jinhu, et al. A review of research on remote sensing for ground impervious surface percentage retrieval[J]. Remote Sensing for Land & Resourses, 2019,31(3):20-28.
14 左家旗,王泽根,边金虎,等.地表不透水面比例遥感反演研究综述[J].国土资源遥感,2019,31(3):20-28.
15 Oon A, Sharfi H Z M, Lechner A M,et al. Discriminating between large-scale oil palm plantations and smallholdings on tropical peatlands using vegetation indices and supervised classification of Landsat-8[J]. International Journal of Remote Sensing,2019,40(19): 7312-7328.
16 Weng Q, Hu X. Medium spatial resolution satellite imagery for estimating and mapping urban impervious surfaces using LSMA and ANN[J]. IEEE Transactions on Geoscience and Remote Sensing, 2008,46(8):2397-2406.
17 Liu Wenya, Yue Anzhi, Ji Jue,et al. Urban green space extraction from GF-2 remote sensing image based on DeepLabv3+ semantic segmentation model[J]. Remote Sensing for Land & Resourses, 2020,32(2):120-129.
17 刘文雅,岳安志,季珏,等.基于DeepLabv3+语义分割模型的GF-2影像城市绿地提取[J].国土资源遥感,2020,32(2):120-129.
18 Luo Qi, Liu Xiaolong, Shi Zhengtao,et al. Study on plastic mulch identification based on the fusion of GF-1 and Sentinel-2 images[J]. Geography and Geo-Information Science, 2021,37(1):39-46.
18 罗琪,刘晓龙,史正涛,等.基于GF-1与Sentinel-2融合数据的地膜识别方法研究[J].地理与地理信息科学,2021,37(1):39-46.
19 Zhang Zerui, Liu Xiaoping, Zhang Honghui, et al. Automatic detection of newly increased construction land based on deep learning and multi-source remote sensing data[J]. Acta Scientiarum Naturalium Universitatis Sunyatseni,2022,61(2):28-37.
19 张泽瑞,刘小平,张鸿辉,等.基于深度学习与多源遥感数据的新增建设用地自动检测[J].中山大学学报(自然科学版),2022,61(2):28-37.
20 Ji Mengyu, Ji Xiaoming, Yu Zhilou. A review of semantic segmentation based on deep learning[J]. Information Technology and Informatization, 2017(10):137-140.
20 计梦予,袭肖明,于治楼.基于深度学习的语义分割方法综述[J].信息技术与信息化,2017(10):137-140.
21 Long J, Shelhamer E, Darrell T. Fully convolutional networks for semantic segmentation[C]∥ Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,2015:3431-3440.
22 Ronnebergero, Fischer P, Brox T. U-Net: Convolutional networks for biomedical image segmentation[C]∥ International Conference on Medical Image Computing and Computer-assisted Intervention, 2015:234-241.
23 Zhao H, Shi J, Qi X, et al. Pyramid Scene Parsing Network[R]. IEEE Computer Society, 2016.
24 Wu Hua, Zhang Xinchang, Sun Ying,et al. Building extraction in complex scenes based on the fusion of multi-feature improved PSPNet model[J]. Bulletin of Surveying and Mapping, 2021(6):21-27.
24 武花,张新长,孙颖,等.融合多特征改进型PSPNet模型应用于复杂场景下的建筑物提取[J].测绘通报,2021(6):21-27.
25 Lu Qi, Qin Jun, Yao Xuedong,et al. Buildings extraction of GF-2 remote sensing image based on multi-layer perception network[J]. Remote Sensing for Land & Resourses, 2021,33(2):75-84.
25 卢麒,秦军,姚雪东,等.基于多层次感知网络的GF-2遥感影像建筑物提取[J].国土资源遥感,2021,33(2):75-84.
26 Song Tingqiang, Liu Tongxin, Zong Da,et al. Research on road extraction method from remote sensing images based on improved U-Net network[J]. Computer Engineering and Applications,2021,57(14):209-216.
26 宋廷强,刘童心,宗达,等.改进U-Net网络的遥感影像道路提取方法研究[J].计算机工程与应用,2021,57(14):209-216.
[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] 王姗姗,蒲英霞,李升峰,李闰洁,李茂华. 利用长时序流域不透水面数据集探索城市增长模式——以秦淮河流域为例[J]. 遥感技术与应用, 2022, 37(3): 739-750.
[3] 符冰雪,张继超,杜文杰,王鹏龙,孙中昶. 一种不透水面精细制图新方法及其在城市SDGs指标评估上的应用[J]. 遥感技术与应用, 2021, 36(6): 1339-1349.
[4] 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.
[5] 高子为,孙伟伟,程朋根,杨刚,孟祥超. 融合高分辨率遥感影像和POI数据的多特征潜在语义信息用于识别城市功能区[J]. 遥感技术与应用, 2021, 36(3): 618-626.
[6] 陈敏,潘佳威,李江杰,徐璐,刘加敏,韩健,陈奕云. 结合VGGNet与Mask R-CNN的高分辨率遥感影像建设用地检测[J]. 遥感技术与应用, 2021, 36(2): 256-264.
[7] 杨玉婷,汤家法,边金虎,李爱农,雷光斌,黄平,蒋梓淳. 加尔各答市地表温度与不透水面比例季相相关性研究[J]. 遥感技术与应用, 2021, 36(1): 79-89.
[8] 沈谦,朱长明,张新,黄巧华,杨程子. 1992~2013我国干旱区城市不透水遥感制图与扩张过程分析[J]. 遥感技术与应用, 2020, 35(5): 1178-1186.
[9] 刘美,杜国明,于凤荣,匡文慧. 哈尔滨城乡梯度建设用地结构变化及不透水面遥感监测分析[J]. 遥感技术与应用, 2020, 35(5): 1206-1217.
[10] 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.
[11] 王卓,闫浩文,禄小敏,冯天文,李亚珍. 一种改进U-Net的高分辨率遥感影像道路提取方法[J]. 遥感技术与应用, 2020, 35(4): 741-748.
[12] 连喜红,祁元,王宏伟,张金龙,杨瑞. 基于面向对象的青海湖环湖区居民地信息自动化提取[J]. 遥感技术与应用, 2020, 35(4): 775-785.
[13] 李智礼,匡文慧,张澍. 近70 a天津主城区城市土地利用/覆盖变化遥感监测与时空分析[J]. 遥感技术与应用, 2020, 35(3): 527-536.
[14] 陈馨,匡文慧. 基于云平台的中哈干旱区典型城市地表覆盖变化遥感监测与比较[J]. 遥感技术与应用, 2020, 35(3): 548-557.
[15] 何浩,刘修国,沈永林. 基于视差的高分辨率遥感影像建筑物变化检测[J]. 遥感技术与应用, 2019, 34(6): 1315-1323.