Please wait a minute...
 
国土资源遥感  2019, Vol. 31 Issue (2): 44-50    DOI: 10.6046/gtzyyg.2019.02.07
  技术方法 本期目录 | 过刊浏览 | 高级检索 |
基于DCNN特征的建筑物震害损毁区域检测
周阳, 张云生(), 陈斯飏, 邹峥嵘, 朱耀晨, 赵芮雪
中南大学地球科学与信息物理学院,长沙 410083
Disaster damage detection in building areas based on DCNN features
Yang ZHOU, Yunsheng ZHANG(), Siyang CHEN, Zhengrong ZOU, Yaochen ZHU, Ruixue ZHAO
School of Geosciences and Inof-Physics, Central South University, Changsha 410083, China
全文: PDF(5779 KB)   HTML  
输出: BibTeX | EndNote (RIS)      
摘要 

为了提高基于高空间分辨率遥感影像的建筑物震害损毁评估精度,引入深度卷积神经网络(deep convolutional neural network,DCNN)模型,提出一种利用DCNN全连接层特征结合支持向量机 (support vector machine,SVM)进行遥感影像建筑物震害损毁区域检测的方法。首先,利用神经网络前馈方式从DCNN全连接层提取训练样本和待检测区域的特征; 然后,基于样本训练SVM分类器; 最后,对待检测区域的所有区块进行分类预测和投票确定是否损毁。以2010年海地地震遥感影像为例,建筑物损毁检测正确率可以达到89%,相比于传统的特征提取方法正确率提高了4%。实验结果表明该方法在建筑物震害损毁检测方面具有一定的应用潜力。

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
周阳
张云生
陈斯飏
邹峥嵘
朱耀晨
赵芮雪
关键词 损毁建筑物卷积神经网络SVM高空间分辨率遥感影像    
Abstract

In order to improve the precision of damage assessment of post-earthquake buildings based on remote sensing images, this paper introduces a deep convolutional neural network (DCNN) model that performs well in natural image classification and target detection, and also proposes a method of using DCNN fully-connected layer features combined with support vector machine (SVM) to detect damaged building areas in remote sensing images. Firstly, neural network feed forward is used to extract the features of the training samples and the regions to be detected from the DCNN fully-connected layer; then the SVM classifier is learned based on the training samples; finally, all the blocks in the detection region are subjected to predicting and voting to determine whether they are damaged. The authors used Haiti earthquake remote sensing imagery in 2010 to do verification. The accuracy rate of damage detection in this method can reach 89%. Compared with the traditional feature extraction method, the correct rate is improved by 4%. The experimental results show that this method has a certain potential in the detection of building damage damage.

Key wordsdamaged building    convolutional neural network    SVM    high-resolution remote sensing image
收稿日期: 2018-04-10      出版日期: 2019-05-23
:  TP751  
基金资助:国家重点研发计划项目“一体化综合减灾智能服务系统”(2016YFC0803108);湖南省自然科学基金项目“基于深度学习的倾斜摄影模型建筑物提取与三维重建”共同资助(2018JJ3637)
通讯作者: 张云生
作者简介: 周 阳(1992-),女,硕士研究生,主要从事数字摄影测量方面的研究。Email: csuzy_smile@163.com。
引用本文:   
周阳, 张云生, 陈斯飏, 邹峥嵘, 朱耀晨, 赵芮雪. 基于DCNN特征的建筑物震害损毁区域检测[J]. 国土资源遥感, 2019, 31(2): 44-50.
Yang ZHOU, Yunsheng ZHANG, Siyang CHEN, Zhengrong ZOU, Yaochen ZHU, Ruixue ZHAO. Disaster damage detection in building areas based on DCNN features. Remote Sensing for Land & Resources, 2019, 31(2): 44-50.
链接本文:  
https://www.gtzyyg.com/CN/10.6046/gtzyyg.2019.02.07      或      https://www.gtzyyg.com/CN/Y2019/V31/I2/44
Fig.1  DCNN结构
Fig.2  方法流程
Fig.3  2010年海地雅克梅勒地区震后航空影像
Fig.4  损毁区域与未损毁区域样本样例
Fig.5  验证集手工标记真实损毁区域
模型 模型精度 测试精度
AlexNet 97.08 88.33
VGGNet 93.50 91.67
BOW 94.25 84.67
Tab.1  SVM训练结果
Fig.6  验证区域分类结果
模型 VGGNet AlexNet
验证区域 验证区域1 验证区域2 验证区域1 验证区域2
实际检测 损毁 未损毁 总计 损毁 未损毁 总计 损毁 未损毁 总计 损毁 未损毁 总计
损毁 861 129 990 718 166 884 855 291 1146 647 158 805
未损毁 150 1 397 1 547 53 663 716 156 1 235 1 391 124 671 795
总计 1 011 1 526 2 537 771 829 1 600 1 011 1 526 2 537 771 829 1 600
漏检率/% 12.8 6.9 15.4 16.1
误检率/% 9.8 20.0 19.1 19.1
正确率/% 89.0 86.3 82.4 82.4
Tab.2  验证集分类结果
[1] Janalipour M, Mohammadzadeh A . Building damage detection using object-based image analysis and ANFIS from high-resolution image (Case study:BAM earthquake,Iran)[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2016,9(5):1937-1945.
doi: 10.1109/JSTARS.2015.2458582
[2] 郭华东, 鹿琳琳, 马建文 , 等. 一种改进的地震灾害倒塌房屋遥感信息自动识别方法[J]. 科学通报, 2009,54(17):2581-2585.
Guo H D, Lu L L, Ma J W , et al. An improved method for automatic recognition of remote sensing information of earthquake disasters with collapsed houses[J]. Chinese Science Bulletin, 2009,54(17):2581-2585.
[3] 叶昕, 王俊, 秦其明 . 基于高分一号卫星遥感图像的建筑物震害损毁检测研究——以2015年尼泊尔M_S8.1地震为例[J]. 地震学报, 2016,38(3):477-485.
Ye X, Wang J, Qin Q M . Building seismic damage detection research based on high-resolution satellite remote sensing images:A case study of Nepal M_S8.1 earthquake in 2015[J]. Journal of Seismology, 2016,38(3):477-485.
[4] Yamazaki F, Suzuki D, Maruyama Y. Detection of damages due to earthquakes using digital aerial images [C]//The 14th World Conference on Earthquake Engineering, 2008.
[5] Ehrlich D, Bielski C. Texture based change detection of built-up on SPOT panchromatic imagery using PCA [C]//Urban Remote Sensing Event.IEEE, 2011: 77-80.
[6] 于海洋, 程钢, 张育民 , 等. 基于LiDAR和航空影像的地震灾害倒塌建筑物信息提取[J]. 国土资源遥感, 2011,23(3):77-81.doi: 10.6046/gtzyyg.2011.03.14.
doi: 10.6046/gtzyyg.2011.03.14
Yu H Y, Cheng G, Zhang Y M , et al. Earthquake disaster collapse building information extraction based on LiDAR and aerial images[J]. Remote Sensing for Land and Resources, 2011,23(3):77-81.doi: 10.6046/gtzyyg.2011.03.14.
[7] 董燕生, 潘耀忠, 方伟华 , 等. 基于面向对象技术的建筑物震害识别方法研究[J]. 地震研究, 2011,34(3):372-377.
Dong Y S, Pan Y Z, Fang W H , et al. Research on building earthquake damage identification method based on object-oriented technology[J]. Earthquake Research, 2011,34(3):372-377.
[8] 王岩, 王晓青, 窦爱霞 . 面向对象遥感分类方法在汶川地震震害提取中的应用[J]. 地震, 2009,29(3):54-60.
Wang Y, Wang X Q, Dou A X . Application of object-oriented remote sensing classification method in seismic damage extraction of Wenchuan earthquake[J]. Earthquake, 2009,29(3):54-60.
[9] 王慧敏, 李艳 . 面向对象的损毁建筑物提取[J].遥感信息, 2011 (5):81-85.
Wang H M, Li Y . Object-oriented damaged building extraction[J].Remote Sensing Information, 2011 (5):81-85.
[10] 刘宇, 曹国, 周丽存 , 等. 基于多特征结合的损毁建筑物检测[J]. 计算机应用, 2015,35(9):2652-2655.
doi: 10.11772/j.issn.1001-9081.2015.09.2652
Liu Y, Cao G, Zhou L C , et al. Damaged building detection based on multi-features combination[J]. Computer Application, 2015,35(9):2652-2655.
[11] 许夙晖, 慕晓冬, 赵鹏 , 等. 利用多尺度特征与深度网络对遥感影像进行场景分类[J]. 测绘学报, 2016,45(7):834-840.
Xu S H, Mu X D, Zhao P , et al. Multi-scale feature and deep network for scene classification of remote sensing images[J]. Journal of Surveying and Mapping, 2016,45(7):834-840.
[12] Hu F, Xia G S, Hu J , et al. Transferring deep convolutional neural networks for the scene classification of high-resolution remote sensing imagery[J]. Remote Sensing, 2015,7(11):14680-14707.
doi: 10.3390/rs71114680
[13] Zhou W, Newsam S, Li C , et al. Learning low dimensional convolutional neural networks for high-resolution remote sensing image retrieval[J]. Remote Sensing, 2017,9(5):489.
doi: 10.3390/rs9050489
[14] Vetrivel A, Gerke M, Kerle N , et al. Disaster damage detection through synergistic use of deep learning and 3D point cloud features derived from very high resolution oblique aerial images,and multiple-kernel-learning[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2017,140:45-59.
[15] 李奇峰 . 结合多特征描述和SVM的遥感影像分类研究[D]. 郑州:郑州大学, 2015.
Li Q F . Research on Remote Sensing Image Classification Based on Multiple Feature Description and SVM[D]. Zhengzhou:Zhengzhou University, 2015.
[16] 王旭东, 段福洲, 屈新原 , 等. 面向对象和SVM结合的无人机数据建筑物提取[J]. 国土资源遥感, 2017,29(1):97-103.doi: 10.6046/gtzyyg.2017.01.15.
Wang X D, Duan F Z, Qu X Y , et al. UAV data architecture extraction combined with object-oriented and SVM[J]. Remote Sensing for Land and Resources, 2017,29(1):97-103.doi: 10.6046/gtzyyg.2017.01.15.
[17] 张峰, 薛艳丽, 李英成 , 等. 基于SVM的多源遥感影像面向对象建筑物提取方法[J]. 国土资源遥感, 2008,20(2):27-29.doi: 10.6046/gtzyyg.2008.02.07.
doi: 10.6046/gtzyyg.2008.02.07
Zhang F, Xue Y L, Li Y C , et al. Multi-source remote sensing image-based object-oriented building extraction method based on SVM[J]. Remote Sensing for Land and Resources, 2008,20(2):27-29.doi: 10.6046/gtzyyg.2008.02.07.
[18] Dalal N, Triggs B. Histograms of oriented gradients for human detection [C]//2005 IEEE Computer Society Conference on Computer Vision and Patter Recognition, 2005: 886-893.
[19] Simonyan K , Zisserman A .Very deep convolutional networks for large-scale image recognition[EB/OL]..
[20] Penatti O A B, Nogueira K, Santos J A D. Do deep features generalize from everyday objects to remote sensing and aerial scenes domains? [C]//Computer Vision and Pattern Recognition Workshops.IEEE, 2015: 44-51.
[21] Wang J, Luo C, Huang H , et al. Transferring pre-trained deep CNNs for remote scene classification with general features learned from linear PCA network[J]. Remote Sensing, 2017,9(3):225.
doi: 10.3390/rs9030225
[22] Cortes C, Vapnik V N . Support-vector networks[J]. Machine Learning, 1995,20(3):273-297.
[23] Xia G S, Hu J, Hu F , et al. AID:A benchmark data set for performance evaluation of aerial scene classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2017,55(7):3965-3981.
doi: 10.1109/TGRS.2017.2685945
[1] 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.
[2] 王华, 李卫卫, 李志刚, 陈学业, 孙乐. 基于多尺度超像素的高光谱图像分类研究[J]. 自然资源遥感, 2021, 33(3): 63-71.
[3] 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.
[4] 仇一帆, 柴登峰. 无人工标注数据的Landsat影像云检测深度学习方法[J]. 国土资源遥感, 2021, 33(1): 102-107.
[5] 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.
[6] 王小龙, 闫浩文, 周亮, 张黎明, 党雪薇. 利用SVM分类Landsat影像的朝鲜主要城市建设用地时空特征分析[J]. 国土资源遥感, 2020, 32(4): 163-171.
[7] 卫虹宇, 赵银娣, 董霁红. 基于改进RetinaNet的冷却塔目标检测[J]. 国土资源遥感, 2020, 32(4): 68-73.
[8] 刘钊, 廖斐凡, 赵桐. 基于PSPNet的遥感影像城市建成区提取及其优化方法[J]. 国土资源遥感, 2020, 32(4): 84-89.
[9] 李宇, 肖春姣, 张洪群, 李湘眷, 陈俊. 深度卷积融合条件随机场的遥感图像语义分割[J]. 国土资源遥感, 2020, 32(3): 15-22.
[10] 蔡之灵, 翁谦, 叶少珍, 简彩仁. 基于Inception-V3模型的高分遥感影像场景分类[J]. 国土资源遥感, 2020, 32(3): 80-89.
[11] 吴同, 彭玲, 胡媛. 基于SU-RetinaNet的高分辨率遥感影像非正规垃圾堆检测[J]. 国土资源遥感, 2020, 32(3): 90-97.
[12] 张春森, 吴蓉蓉, 李国君, 崔卫红, 冯晨轶. 面向对象的高空间分辨率遥感影像箱线图变化检测方法[J]. 国土资源遥感, 2020, 32(2): 19-25.
[13] 娄佩卿, 陈晓雨, 王疏桐, 付波霖, 黄永怡, 唐廷元, 凌铭. 基于无人机影像的喀斯特农耕区地物识别——以桂林市为例[J]. 国土资源遥感, 2020, 32(1): 216-223.
[14] 姚丙秀, 黄亮, 许艳松. 一种结合超像素和图论的高空间分辨率遥感影像分割方法[J]. 国土资源遥感, 2019, 31(3): 72-79.
[15] 叶发茂, 罗威, 苏燕飞, 赵旭青, 肖慧, 闵卫东. 卷积神经网络特征在遥感图像配准中的应用[J]. 国土资源遥感, 2019, 31(2): 32-37.
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