基于特征增强和ELU的神经网络建筑物提取研究

doi:10.12082/dqxxkx.2021.200130

地球信息科学学报 ›› 2021, Vol. 23 ›› Issue (4): 692-709.doi: 10.12082/dqxxkx.2021.200130

基于特征增强和ELU的神经网络建筑物提取研究

唐璎¹^,²^,³^,⁴, 刘正军¹^,^*, 杨懿¹, 顾海燕¹, 杨树文²^,³^,⁴

1.中国测绘科学研究院摄影测量与遥感研究所,北京 100830
2.兰州交通大学测绘与地理信息学院,兰州 730070
3.地理国情监测技术应用国家地方联合工程研究中心,兰州 730070
4.甘肃省地理国情监测工程实验室,兰州 730070

收稿日期:2020-03-21 修回日期:2020-06-21 出版日期:2021-04-25 发布日期:2021-06-25
通讯作者: *刘正军（1974— ）,男,湖南湘潭人,研究员,主要从事遥感影像信息提取与生态环境遥感监测、突发事件应急地理信息技术等研究。E-mail: zjliu@casm.ac.cn
基金资助:
国家重点研发计划项目(2018YFB0504504);国家自然科学基金项目(41701506、41371406);中央级公益性科研院所基本科研业务费专项资金项目(AR1923)

Research on Building Extraction based on Neural Network with Feature Enhancement and ELU Activation Function

TANG Ying¹^,²^,³^,⁴, LIU Zhengjun¹^,^*, YANG Yi¹, GU Haiyan¹, YANG Shuwen²^,³^,⁴

1. Institute of Photogrammetry and Remote Sensing, Chinese Academy of Surveying and Mapping, Beijing 100830, China
2. Faculty of Geomatics, Lanzhou Jiaotong University, Lanzhou 730070, China
3. National-Local Joint Engineering Research Center of Technologies and Applications for National Geographic State Monitoring, Lanzhou 730070, China
4. Gansu Provincial Engineering Laboratory for National Geographic State Monitoring, Lanzhou 730070, China

Received:2020-03-21 Revised:2020-06-21 Online:2021-04-25 Published:2021-06-25
Supported by:
The National Key R&D Program of China, No.2018YFB0504504(2018YFB0504504);N-ational Natural Science Foundation of China, No.41701506, 41371406(41701506、41371406);Central Public-interest Scientific Institution Basal Research Fund, No.AR1923(AR1923)

摘要/Abstract

摘要：

近年来,城市发展快速,大量人口奔向城市工作生活,城市建筑物的数量有如雨后春笋般扩张,需要合理地规划城市土地资源,遏制违规乱建现象,因此基于高分辨率遥感影像,对建筑物进行准确提取,对城市规划和管理有着重要辅助作用。本文基于U-Net网络模型,使用美国马萨诸塞州建筑物数据集,对网络模型结构进行探究,提出了一种激活函数为ELU、“编码器-特征增强-解码器”结构的网络模型FE-Net。实验首先通过比较不同网络层数的U-Net5、U-Net6、U-Net7的建筑物提取效果,找到最佳的基础网络模型U-Net6;其次,基于该模型,加入特征增强结构得到“U-Net6+ReLU+特征增强”的网络模型;最后,考虑到ReLU容易产生神经元死亡,为优化激活函数,将激活函数替换为ELU,从而得到网络模型FE-Net（U-Net6+ELU+特征增强）。比较3个网络模型（U-Net6+ReLU、U-Net6+ReLU+特征增强、FE-Net（U-Net6+ELU+特征增强））的建筑物提取结果,表明FE-Net网络模型的建筑物提取效果最好,精度放松F1值达到97.23%,比“U-Net6+ReLU”和“U-Net6+ReLU+特征增强”2个网络模型分别高出0.36%和0.12%,且与其他具有相同数据集的研究成果比较,具有最高的提取精度,它能较好地提取出多尺度的建筑物,不仅对小尺度建筑物有较好的提取效果,而且能大致、较完整地提取出形状不规则的建筑物,有相对更少的漏检和错检,较准确地实现了端到端的建筑物提取。

关键词: 高分辨率遥感影像, 卷积神经网络, 建筑物提取, 特征增强, 激活函数ELU, FE-Net网络模型, 端到端, 深度学习

Abstract:

In recent years, with the rapid development of the city, a large number of people turn to work and live in the city, resulting in an increasing number of urban buildings. Land resources and urban ecological environment (such as green space) are threatened to some extent. Thus, it is urgent to plan urban land resources and space reasonably, prevent illegal construction, improve urban living environment, and make the city sustainable, orderly, healthy, and green. With the high-resolution remote sensing image data becoming more and more abundant, accurate building extraction using high-resolution remote sensing images plays an important role in urban planning, urban management, and change detection of urban buildings. Based on the U-Net network model, using the Massachusetts building dataset, this paper explored the network model structure and proposed a network model called FE-Net with "encoder-feature enhancement-decoder" structure and ELU activation function. First, the best basic network model called U-Net6 was found by comparing the building extraction results using U-Net5, U-Net6, and U-Net7 with different number of network layers. Based on the U-Net6, the network model of "U-Net6+ReLU+feature enhancement" was established by adding the structure of feature enhancement. In order to optimize the activation function, the ReLU activation function was replaced by the ELU activation function, and then the network model called FE-Net (U-Net6+ELU+feature enhancement) was created. The FE-Net network model was compared with the building extraction results from the other two network models (U-Net6+ReLU and U-Net6+ReLU+feature enhancement). Results show that the FE-Net network model had the best building extraction performance. Its relaxed F1-measure reached 97.23%, which was 0.36% and 0.12% higher than the other two network models. Meanwhile, FE-Net also had the highest extraction accuracy compared with other studies using the same dataset of Massachusetts. The FE-Net network model can extract multi-scale buildings better, which can not only extract small-scale buildings accurately, but also roughly and completely extract buildings with irregular shape with relatively less missing and wrong detections. Thus, the FE-Net network model can be used to achieve end-to-end building extraction with a high accuracy.

Key words: high-resolution remote sensing image, convolutional neural network, building extraction, feature enhancement, Exponential Linear Units (ELU), Feature Enhancement Network (FE-Net), end-to-end, deep learning

唐璎, 刘正军, 杨懿, 顾海燕, 杨树文. 基于特征增强和ELU的神经网络建筑物提取研究[J]. 地球信息科学学报, 2021, 23(4): 692-709.DOI:10.12082/dqxxkx.2021.200130

TANG Ying, LIU Zhengjun, YANG Yi, GU Haiyan, YANG Shuwen. Research on Building Extraction based on Neural Network with Feature Enhancement and ELU Activation Function[J]. Journal of Geo-information Science, 2021, 23(4): 692-709.DOI:10.12082/dqxxkx.2021.200130

图/表 19

图1

图2

图3

图4

图5

图6

图7

图8

表1

图9

表2

图10

图11

表3

图12

图13

图14

表4

表5

参考文献 30

[1]	Han J W, Zhang D W, Cheng G, et al. Object detection in optical remote sensing images based on weakly supervised learning and high-level feature learning[J]. IEEE Transactions on Geoscience and Remote Sensing, 2015,53(6):3325-3337.
[2]	黄金库, 冯险峰, 徐秀莉, 等. 基于知识规则构建和形态学修复的建筑物提取研究[J]. 地理与地理信息科学, 2011,27(4):28-31.
	[ Huang J K, Feng X F, Xu X L, et al. Research on building extraction based on knowledge rule construction and morphological restoration[J]. Geography and Geo-Information Science, 2011,27(4):28-31. ]
[3]	林雨准, 张保明, 徐俊峰, 等. 多特征多尺度相结合的高分辨率遥感影像建筑物提取[J]. 测绘通报, 2017(12):53-57.
	[ Lin Y Z, Zhang B M, Xu J F, et al. High resolution remote sensing image building extraction based on multi feature and multi-scale[J]. Bulletin of Surveying and Mapping, 2017(12):53-57. ]
[4]	林祥国, 张继贤. 面向对象的形态学建筑物指数及其高分辨率遥感影像建筑物提取应用[J]. 测绘学报, 2017,46(6):724-733.
	[ Lin X G, Zhang J X. Object oriented morphological building index and its application in building extraction from high-resolution remote sensing image[J]. Acta Geodaetica et Cartographica Sinica, 2017,46(6):724-733. ]
[5]	田昊, 杨剑, 汪彦明, 等. 基于先验形状约束水平集模型的建筑物提取方法[J]. 自动化学报, 2010,36(11):1502-1511.
	[ Tian H, Yang J, Wang Y M, et al. Building extraction method based on prior shape constraint level set model[J]. Acta Automatica Sinica, 2010,36(11):1502-1511. ]
[6]	黄昕. 高分辨率遥感影像多尺度纹理、形状特征提取与面向对象分类研究[D]. 武汉:武汉大学, 2009.
	[ Huang X. Research on multi-scale texture, shape feature extraction and object-oriented classification of high-resolution remote sensing image[D]. Wuhan: Wuhan University, 2009. ]
[7]	Levitt S, Aghdasi F. Texture measures for building recognition in aerial photographs[J]. Communications and Singnal Processing, 1997:75-80.
[8]	Lin C. Detection of buildings using perceptual grouping and shadows[J]. IEEE Computer Vision & Pattern Recognition, 1994:62-69.
[9]	Kim T, Muller J P. Development of a graph-based approach for building detection[J]. Image and Vision Computing, 1999,17(1):3-14.
[10]	Katartzis A, Sahli H, Nyssen E, et al. Detection of buildings from a single airborne image using a markov random field model[J]. Geoscience and Remote Sensing Symposium, 2001,6:2832-2834.
[11]	Jung C R, Schramm R. Rectangle detection based on a windowed hough transform[J]. Computer Graphics and Image Processing, 2004:113-120.
[12]	Mnih V. Machine learning for aerial image labeling[D]. Toronto: University of Toronto, 2013.
[13]	Saito S, Yamashita Y, Aoki Y. Multiple object extraction from aerial imagery with convolutional neural networks[J]. Journal of Imaging Science & Technology, 2016,60(1):10402-1/10402-9. doi: 10.2352/J.ImagingSci.Technol.(2008)52:4(040502) pmid: 19756243
[14]	Saito S, Aoki Y. Building and road detection from large aerial imagery[C]. Proceedings of SPIE/IS&T Electronics Imaging. International Society for Optics and Photonics, 2015:94050K-94050K.
[15]	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.
	[ Liu W T, Li S H, Qin Y C. Automatic building roof extraction with fully convolutional neural network[J]. Journal of Geo-information Science, 2018,20(11):1562-1570. ]
[16]	Xu Y Y, Wu L, Xie Z, et al. Building extraction in very high resolution remote sensing imagery using deep learning and guided filters[J]. Remote Sensing, 2018,10:144-161.
[17]	Qin Y C, Wu Y C, Li B, et al. Semantic segmentation of building roof in dense urban environme-nt with deep convolutional neural network: A case study using GF2 VHR imagery in China[J]. Sensors, 2019,19:1164-1175.
[18]	Ye Z R, Fu Y Y, Gan M Y, et al. Building extraction from very high resolution aerial imagery using joint attention deep neural network[J]. Remote Sensing, 2019,11:2970-2990.
[19]	王宇, 杨艺, 王宝山, 等. 深度神经网络条件随机场高分辨率遥感图像建筑物分割[J]. 遥感学报, 2019,23(6):1194-1208.
	[ Wang Y, Yang Y, Wang B S, et al. Building segmentation of high-resolution remote sensing image of random airport under the condition of deep neural network[J]. Journal of Remote Sensing, 2019,23(6):1194-1208 ]
[20]	范荣双, 陈洋, 徐启恒, 等. 基于深度学习的高分辨率遥感影像建筑物提取方法[J]. 测绘学报, 2019,48(1):34-41.
	[ Fan R S, Chen Y, Xu Q H, et al. Building extraction method of high-resolution remote sensing image based on deep learning[J]. Acta Geodaetica et Cartographica Sinica, 2019,48(1):34-41. ]
[21]	Liu Y H, Gross L, Li Z Q, et al. Automatic building extraxtion on high-resolution remote sensing i-magery using deep convolutional encoder-decoder with spatial pyramid pooling[J]. IEEE Access, 2019,7:128774-128786.
[22]	杨嘉树, 梅天灿, 仲思东. 顾及局部特性的 CNN 在遥感影像分类的应用[J]. 计算机工程与应用, 2018,54(7):188-195.
	[ Yang J S, Mei T C, Zhong S D. Application of CNN considering local characteristics in remote sensing image classification[J]. Computer Engineering and Application, 2018,54(7):188-195. ]
[23]	Ronneberger O, Fischer P, Brox T. U-net: convolutional networks for biomedical image segmentation[C]. International Conference on Medial Image Computing and Computer-Assisted Intervention, Cham: Springer International Publishing, 2015:234-241.
[24]	Clevert D A, Unterthiner T, Hochreiter S. Fast and accurate deep network learning by exponential linear units (elus)[J]. Computer Science, 2015:334-337.
[25]	Nair V, Hinton G E. Rectified linear units improve restricted boltzmann machines[C]. International Conference on Machine Learning, 2010.
[26]	Zhou L C, Zhang C, Wu M. D-LinkNet: LinkNet with pretrained encoder and dilated convolution for high resolution satellite imagery road extraction[C]. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) IEEE, 2018:182-186.
[27]	Wiedemann C, Heipke C, Mayer H. Empirical evaluation of automatically extracted road axes[J]. Empirical Evaluation Techniques in Computer Vision, 1998:172-187.
[28]	Marcu A, Costea D, Slusanschi E, et al. A multi-stage multi-task neural network for aerial scene in-terpretation and geolocalization[J]. ArXiv:1804.01322v1, 2018.
[29]	Khalel A, El-Saban M. Automatic pixelwise object labeling for aerial imagery using stacked u-nets[J]. ArXiv:1803.04953, 2018.
[30]	Pan X, Yang F, Gao L, et al. Building extraction from high-resolution aerial imagery using a generative adversarial network with spatial and channel attention mechanisms[J]. Remote Sensing, 2019,11(8):917-934.

编码器部分		解码器部分
名称	尺寸和通道数	名称	尺寸和通道数
输入影像	384×384×3	特征图	12×12×1024
Conv(3×3), 32	384×384×32	Transposed Conv(3×3), m×2+拼接	24×24×1024
Conv(3×3), n×1	384×384×32	Conv(3×3), n×(1/2)	24×24×512
Maxpool(2×2), m×(1/2)	192×192×32	Conv(3×3), n×1	24×24×512
Conv(3×3), n×2	192×192×64	Transposed Conv(3×3), m×2+拼接	48×48×512
Conv(3×3), n×1	192×192×64	Conv(3×3), n×(1/2)	48×48×256
Maxpool(2×2), m×(1/2)	96×96×64	Conv(3×3), n×1	48×48×256
Conv(3×3), n×2	96×96×128	Transposed Conv(3×3), m×2+拼接	96×96×256
Conv(3×3), n×1	96×96×128	Conv(3×3), n×(1/2)	96×96×128
Maxpool(2×2), m×(1/2)	48×48×128	Conv(3×3), n×1	96×96×128
Conv(3×3), n×2	48×48×256	Transposed Conv(3×3), m×2+拼接	192×192×128
Conv(3×3), n×1	48×48×256	Conv(3×3), n×(1/2)	192×192×64
Maxpool(2×2), m×(1/2)	24×24×256	Conv(3×3), n×1	192×192×64
Conv(3×3), n×2	24×24×512	Transposed Conv(3×3), m×2+拼接	384×384×64
Conv(3×3), n×1	24×24×512	Conv(3×3), n×(1/2)	384×384×32
Maxpool(2×2), m×(1/2)	12×12×512	Conv(3×3), n×1	384×384×32
Conv(3×3), n×2	12×12×1024	Conv(3×3), 1	384×384×1
Conv(3×3), n×1	12×12×1024

项目	参数
中央处理器	Intel® Xeon(R) CPU E5-2620 v2 @2.10GHz × 24
内存	62.8 GB
硬盘	2 TB
显卡	GeForce GTX 1080/PCIe/SSE2
操作系统	Ubuntu 16.04
开发语言	Python
深度学习框架	PyTorch

模型	网络层数	放松精准率（ρ=3）/%	放松召回率（ρ=3）/%	放松F1值（ρ=3）/%	训练时间/s
U-Net5	5	95.43	92.97	94.10	85 761
U-Net6	6	98.31	95.46	96.87	119 610
U-Net7	7	96.58	91.33	93.89	193 152
U-Net8	8	训练崩溃,训练时间太长

模型	放松精准率（ρ=3）/%	放松召回率（ρ=3）/%	放松F1值（ρ=3）/%	训练时间/s
U-Net6+ ReLU	98.31	95.46	96.87	119 610
U-Net6+ ReLU+特征增强	98.50	95.93	97.11	136 487
U-Net6+ ELU+特征增强（FE-Net）	98.59	96.11	97.23	138 135

方法	年份	放松F1值(ρ=3)/%
Mnih^[12]	2013	92.11
Marcu等^[28]	2018	96.04
Khalel等^[29]	2018	96.33
Pan等^[30]	2019	96.36
Ours	2020	97.23

基于特征增强和ELU的神经网络建筑物提取研究

Research on Building Extraction based on Neural Network with Feature Enhancement and ELU Activation Function

RichHTML

PDF (PC)

赞

可视化

摘要/Abstract

引用本文

使用本文

图/表 19

参考文献 30

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]	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.
[12]	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.
[13]	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.
[14]	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.
[15]	于明洋, 陈肖娴, 张文焯, 刘耀辉. 融合网格注意力阀门和特征金字塔结构的高分辨率遥感影像建筑物提取[J]. 地球信息科学学报, 2022, 24(9): 1785-1802.