基于全卷积神经网络的建筑物屋顶自动提取

doi:10.12082/dqxxkx.2018.180159

地球信息科学学报 ›› 2018, Vol. 20 ›› Issue (11): 1562-1570.doi: 10.12082/dqxxkx.2018.180159

• 地球信息科学理论与方法 • 上一篇下一篇

基于全卷积神经网络的建筑物屋顶自动提取

刘文涛¹(), 李世华¹, 覃驭楚^2,^*()

1. 电子科技大学资源与环境学院,成都 611731
2. 中国科学院遥感与数字地球研究所遥感科学国家重点实验室,北京 100101

收稿日期:2018-04-02 修回日期:2018-09-14 出版日期:2018-11-20 发布日期:2018-11-20
作者简介:
作者简介：刘文涛（1989- ）,男,硕士,主要研究方向为计算机视觉。E-mail: liuwentaoboy@126.com
基金资助:
中国科学院百人计划项目(Y6YR0700QM);国家自然科学基金项目(41471294)

Automatic Building Roof Extraction with Fully Convolutional Neural Network

LIU Wentao¹(), LI Shihua¹, QIN Yuchu^2,^*()

1. School of Resources and Environment, University of Electronic Science and Technology of China, Chengdu 611731, China
2. State Key Laboratory of Remote Sensing Science, Institute of Remote Sensing and Digital Earth, Chinese Academy of Science, Beijing 100101, China

Received:2018-04-02 Revised:2018-09-14 Online:2018-11-20 Published:2018-11-20
Contact: QIN Yuchu
Supported by:
100 Talents Program of the Chinese Academy of Sciences, No.Y6YR0700QM;National Natural Science Foundation of China, No.41471294.

摘要/Abstract

摘要：

高分辨率遥感影像在地面自动目标提取中得到了广泛应用,然而利用传统算法,很难高精度地进行实时的建筑物屋顶绘图。本文使用深度学习方法探讨建筑物屋顶分割,由于卷积运算对形变、旋转、光照条件的不敏感,设计了一种用于建筑物屋顶提取的深度卷积神经网络,提出的网络为级联式全卷积神经网络,在深度卷积神经网络的设计中使用了特征复用和特征增强,实现建筑物的自动精确提取。以美国马萨诸塞州建筑物数据集为基础的实验结果表明,本文提出的网络结构取得了92.3%的总体预测精度,和其他方法相比,本文提出的方法具有更高的精度

关键词: 遥感图像, 建筑物, 深度学习, 卷积神经网络, 自动提取

Abstract:

The very high resolution remotely sensed imagery has been widely applied in automatic extraction of ground objects, However, it is hard to conduct timely operational building roof mapping with high accuracy using conventional algorithms. This paper investigate building roof segmentation with deep learning method, since the convolution operations upon images is not sensitive to deformation,rotation and illumination condition, a Deep Convolutional Neural Network (DCNN) is designed for building roof extraction, the proposed network has a cascaded structure with fully convolutional layers, with strategies for feature reuse and enhancement in the design of DCNN, it is expected to accurately extract building roof. The experiment is carried out upon building sample data set acquired in Massachusetts, USA, the results show that the proposed network achieved overall accuracy of 92.3%, and the comparison with other methods suggest the proposed network is able to map building roof with high accuracy.

Key words: remote sensing image, building, deep learning, convolutional neural network, automatic extraction

刘文涛, 李世华, 覃驭楚. 基于全卷积神经网络的建筑物屋顶自动提取[J]. 地球信息科学学报, 2018, 20(11): 1562-1570.DOI:10.12082/dqxxkx.2018.180159

LIU Wentao,LI Shihua,QIN Yuchu. Automatic Building Roof Extraction with Fully Convolutional Neural Network[J]. Journal of Geo-information Science, 2018, 20(11): 1562-1570.DOI:10.12082/dqxxkx.2018.180159

图/表 13

图1

表1

表2

表3

图2

图3

图4

图5

图6

图7

表4

表5

图8

参考文献 23

[1]	严岩. 高空间分辨率遥感影像建筑物提取研究综述[J]. 数字技术与应用, 2012(7):75-78.
	[ Yan Y.Research review on the extraction of high spatial resolution remote sensing images[J]. Digital Technology and Application, 2012(7):75-78. ]
[2]	安文,杨俊峰,赵羲,等.高分辨率遥感影像的建筑物自动提取[J].测绘科学,2014,39(11):80-84.
	[ An W, Yang J F, Zhao Xi, et al.Building automatic extraction of high resolution remote sensing images[J]. Science of Surveying and Mapping, 2014,39(11):80-84. ]
[3]	赵凌君. 高分辨率SAR图像建筑物提取方法研究[D].长沙:国防科学技术大学,2009.
	[ Zhao L J.Research on extraction method of high resolution SAR image buildings[D]. Changsha: National Defense Science and Technology University, 2009. ]
[4]	候蕾,尹东,尤晓建.一种遥感图像中建筑物的自动提取方法[J].计算机仿真,2006,23(4):184-187.
	[ Hou L, Yin D, You X J.The automatic extraction method of buildings in a remote sensing image[J]. Computer simulation, 2006,23(4):184-187. ]
[5]	巩丹超,张永生.基于航空影像的建筑物半自动提取技术研究[J]. 测绘通报, 2002(10):15-18. doi: 10.3969/j.issn.0494-0911.2002.10.005
	[ Gong D C, Zhang Y S.Research on the semi-automatic extraction of buildings based on aerial images[J]. Surveying and mapping, 2002(10):15-18. ] doi: 10.3969/j.issn.0494-0911.2002.10.005
[6]	Bottou L, Bengio Y, Lecun Y.Global training of document processing systems using graph transformer networks[J]. Computer Vision and Pattern Recognition, 1997,6:564-579. doi: 10.1109/CVPR.1997.609370
[7]	Hinton G E, Simon O, Yee-Whye T.A fast learning algorithm for deep belief nets[J]. Neural computation, 2006,18(7):1527-1554. doi: 10.1162/neco.2006.18.7.1527
[8]	Krizhevsky A, Sutskever I, Hinton G E.ImageNet classification with deep convolutional neural networks[J]. Advances in Neural Information Processing Systems, 2012,25(2):12-20. doi: 10.1145/3065386
[9]	曲景影,孙显,高鑫.基于CNN模型的高分辨率遥感图像目标识别[J].研究与开发,2016,36(8):46-51. doi: 10.3969/j.issn.1002-8978.2016.08.011
	[ Qu J Y, Sun X, Gao X.The target recognition of high-resolution remote sensing image based on CNN model[J]. Research and Development, 2016,36(8):46-51. ] doi: 10.3969/j.issn.1002-8978.2016.08.011
[10]	Mnih V.Machine learning for aerial image labeling[D]. Toronto: University of Toronto, 2013.
[11]	Audebert N, Bertrand L S, Lefèvre S.Semantic segmentation of earth observation data using multimodal and multi-scale deep networks[J]. Asian Conference on Computer Vision 2016,9:180-196.
[12]	Saito S, Yamashita T, Aoki Y. Multiple object extraction from aerial imagery with convolutional neural networks[J]. Journal of Imaging Science and Technology, 2015,60(1):10402-1-10402-9. doi: 10.2352/ISSN.2470-1173.2016.10.ROBVIS-392
[13]	Fahlman S E, Lebiere C.The cascade-correlation learning architecture[J]. Neural Information Proccesing Systems, 1989,6:120-125.
[14]	Long J, Shelhamer E, Darrell T.Fully convolutional networks for semantic segmentation[C]. IEEE Conference on Computer Vision and Pattern Recognition, Boston, 2015:3431-3440.
[15]	ChenL C, Jonathan B T, George P, et al. Deep lab semantic image segmentation with task-Specific edge detection using CNNs and a discriminatively trained domain transform[C]. Conference on Computer Vision and Pattern Recognition, Haifa, 2016:834-848.
[16]	Cortes C, Vapnik V.Support-vector networks[J]. Machine Learning, 1995,20(3):273-297.
[17]	Yoav Freund, Robert Schapire E.A Decision-theoretic generalization of on-line learning and an application to boosting[J]. Journal of Computer and System Sciences,1997,55(1):119-139. doi: 10.1007/3-540-59119-2_166
[18]	Lowe, David G.Object recognition from local scale-invariant features[C]. Proceedings of the International Conference on Computer, 1999:1150-1157.
[19]	Simonyan K, Zisserman A.Very deep convolutional networks for large-scale image recognition[J]. Computer Science,2015,2:1409-1556.
[20]	Jia Y, Shelhamer E, Donahue J, et al.Caffe: Convolutional architecture for fast feature embedding[M]. Arxiv: Eprint Arxiv, 2014:675-678.
[21]	Zeiler M D, Fergus R.Visualizing and Understanding Convolutional Networks[C]. Computer Vision-ECCV. Lecture Notes in Computer Science, Cham: Springer, 2014,818-833.
[22]	Boureau Y, Ponce J, LeCun Y. A Theoretical Analysis of Feature Pooling in Visual Recognition[C]. Haifa: International Conference on Machine Learning, 2010.
[23]	美国马萨诸塞州道路与建筑物官方数据集下载网址:

网络层	大小	网络层	大小
Conv1-1	3×3×64	Deconv1-1	1×1
Conv1-2	3×3×64	Deconv1-2	1×1
MaxPooling	2×2	-	-
Conv2-1	3×3×128	Deconv2-1	4×4
Conv2-2	3×3×128	Deconv2-2	4×4
MaxPooling	2×2	-	-
Conv3-1	3×3×256	Deconv3-1	8×8
Conv3-2	3×3×256	Deconv3-2	8×8
Conv3-3	3×3×256	Deconv3-3	8×8
MaxPooling	2×2	-	-
Conv4-1	3×3×512	Deconv4-1	16×16
Conv4-2	3×3×512	Deconv4-2	16×16
Conv4-3	3×3×512	Deconv4-3	16×16
MaxPooling	2×2	-	-
Conv5-1	3×3×512	Deconv5-1	32×32
Conv5-2	3×3×512	Deconv5-2	32×32
Conv5-3	3×3×512	Deconv5-3	32×32

超参数	设置方案一	设置方案二
批处理大小	24	12
权值衰减	5e-4	5e-4
动量因子	0.9	0.9
最大迭代次数	20 000	15 000
学习率	1e-4	1e-5

	实际正类	实际负类
预测正类	TP	FP
预测负类	FN	TN

图像ID	坐标（X）	坐标（Y）
Patch ID：1	600	600
Patch ID：2	700	300
Patch ID：3	900	900
Patch ID：4	500	600
Patch ID：5	600	500

图像ID	Patch ID：1	Patch ID：2	Patch ID：3	Patch ID：4	Patch ID：5
Mnih^[10]	0.890	0.845	0.856	0.863	0.883
Satio^[12]	0.909	0.879	0.899	0.904	0.910
Ours	0.939	0.913	0.934	0.913	0.953
Ours（vs）Mnih^[10]	5.5%+	8.0%+	9.1%+	5.8%+	7.9%+
Ours（vs）Satio^[12]	3.3%+	3.9%+	3.9%+	0.1%+	4.7%+

基于全卷积神经网络的建筑物屋顶自动提取

Automatic Building Roof Extraction with Fully Convolutional Neural Network

RichHTML

PDF (PC)

赞

可视化

被引次数

摘要/Abstract

引用本文

使用本文

图/表 13

参考文献 23

相关文章 15

Metrics

本文评价

推荐阅读 0

[1]	刘潇, 刘智, 林雨准, 王淑香, 左溪冰. 面向遥感影像场景分类的类中心知识蒸馏方法[J]. 地球信息科学学报, 2023, 25(5): 1050-1063.
[2]	侯慧太, 蓝朝桢, 徐青. 基于卫星影像全局和局部深度学习特征检索的无人机绝对定位方法[J]. 地球信息科学学报, 2023, 25(5): 1064-1074.
[3]	陈科, 管海燕, 雷相达, 曹爽. 基于特征增强核点卷积网络的多光谱LiDAR点云分类方法[J]. 地球信息科学学报, 2023, 25(5): 1075-1087.
[4]	张金雷, 陈奕洁, Panchamy Krishnakumari, 金广垠, 王骋程, 杨立兴. 基于注意力机制的城市轨道交通网络级多步短时客流时空综合预测模型[J]. 地球信息科学学报, 2023, 25(4): 698-713.
[5]	衡雪彪, 许捍卫, 唐璐, 汤恒, 许怡蕾. 基于改进全卷积神经网络模型的土地覆盖分类方法研究[J]. 地球信息科学学报, 2023, 25(3): 495-509.
[6]	于方圆, 曹家玮, 李发源, 李思进. 顾及对象特征的地面式光伏电站提取及减碳效益评估[J]. 地球信息科学学报, 2023, 25(3): 529-545.
[7]	徐繁树, 王保云, 韩俊. 一种沟谷型潜在泥石流危险性评价方法：基于多源数据融合的卷积神经网络[J]. 地球信息科学学报, 2023, 25(3): 588-605.
[8]	高建文, 管海燕, 彭代锋, 许正森, 康健, 季雅婷, 翟若雪. 基于局部-全局语义特征增强的遥感影像变化检测网络模型[J]. 地球信息科学学报, 2023, 25(3): 625-637.
[9]	高鹏飞, 曹雪峰, 李科, 游雄. 融合多元稀疏特征与阶层深度特征的遥感影像目标检测[J]. 地球信息科学学报, 2023, 25(3): 638-653.
[10]	黄帅元, 董有福, 李海鹏. 黄土高原区SRTM1 DEM高程误差校正模型构建及对比分析[J]. 地球信息科学学报, 2023, 25(3): 669-681.
[11]	马京振, 周炤, 张付兵. 顾及多特征约束的建筑物组合化简方法[J]. 地球信息科学学报, 2023, 25(2): 277-287.
[12]	江宝得, 许少芬, 王进, 王淼. 基于条件生成对抗网络的电子地图可视水印去除方法[J]. 地球信息科学学报, 2023, 25(2): 288-297.
[13]	谢谦, 陆明, 谢春山. 基于LBS和深度学习的旅游景区客流量的高时频预测[J]. 地球信息科学学报, 2023, 25(2): 298-310.
[14]	黄昕, 毛政元. 基于时空多图卷积网络的网约车乘客需求预测[J]. 地球信息科学学报, 2023, 25(2): 311-323.
[15]	刘洋, 康健, 管海燕, 汪汉云. 基于双注意力残差网络的高分遥感影像道路提取模型[J]. 地球信息科学学报, 2023, 25(2): 396-408.