Detecting Building Façade Damage Caused by Earthquake Using CBAM-Improved Mask R-CNN
-
摘要: 震后建筑物损毁信息是灾情快速评估和应急救援的重要决策依据之一。针对传统的建筑物损毁遥感检测技术只关注于顶面信息，导致众多顶面结构完好而中间层、底层倒塌或崩裂的损毁建筑物处于检测盲点的问题，提出了一种融合引入注意力机制的深度学习实例分割模型和图像多尺度分割算法进行震后建筑物立面损毁检测的方法。首先，利用卷积块状注意力模块（convolutional block attention module, CBAM）改进掩模区域卷积神经网络（mask region-based convolutional networks, Mask R-CNN)模型实现了复杂建筑物立面背景中的损毁信息提取；然后，基于建筑物立面影像多尺度分割结果，利用多数投票规则实现了损毁检测结果的后处理优化。实验结果表明，相比传统损毁检测方法，所提方法能够更有效地实现震后建筑物立面损毁信息的精准定位，总体准确率可达到89.15%。Abstract:Objectives Building damage information can provide an important basis for the decision making of rapid post-earthquake assessment. Traditional building damage detection techniques mainly focus on the roof surface, thus many damaged buildings with an intact roof surface but collapsed middle floors may be neglected. We propose a method of building façade damage detection based on deep learning and multiresolution segmentation algorithm.Methods The method which integrates the instance segmentation and multiresolution segmentation algorithm is applied to detecting the post-earthquake building façade damage. The first thing is to collect the ground images of post-earthquake buildings in the field and perform the data augmentation. Secondly, we use the convolutional block attention module (CBAM) to improve Mask R-CNN. Then the dataset is input to the improved model for training, and finally a multiresolution segmentation algorithm is adopted to post-process the building façade damage detection results output by the CBAM-Improved Mask R-CNN.Results The experimental results show: (1) Collecting ground images of buildings in the field and performing image augmentation can effectively guarantee the necessary training sample size of the instance segmentation model. (2) The Mask R-CNN improved by CBAM attention mechanism significantly improves the post-earthquake building facade damage detection capabilities, which realizes the precise extraction of damage information from complex building façade backgrounds. (3) In addition, using the multiresolution segmentation algorithm to post-process the building facade damage detection results can obviously solve the blurred boundary problems caused by the accumulation of convolutional layers.Conclusions The proposed method can significantly improve the capability of post-earthquake building façade damage detection when compared to the traditional methods, which also raises the Mask R-CNN's accuracy, precision, recall and F2-score to a certain degree. It can be inferred that the proposed method has the strong potential to be applied to the post-earthquake building façade damage detection and therefore provides an important technical means for detecting the comprehensive and detailed building damage detection caused by earthquake.
-
-
![]()
图 1 震后顶面完好而立面破损的建筑物
Figure 1. Post-Earthquake Building with an Intact Roof Surface and Damaged Façade
![]()
图 2 震后建筑物立面损毁检测流程图
Figure 2. Flowchart of Post-Earthquake Building Facade Damage Detection
![]()
图 5 引入CBAM注意力机制的Mask R-CNN模型特征提取网络
Figure 5. Feature Extraction Network of Mask R-CNN Model Improved by CBAM Attention Mechanism
![]()
图 8 CMR模型输出北川遗址建筑物立面损毁检测结果
Figure 8. Beichuan Building Façade Damage Detection Results Using CMR Model
![]()
图 9 CMR模型输出汉旺遗址建筑物立面损毁检测结果
Figure 9. Hanwang Building Façade Damage Detection Results Using CMR Model
表 1 精度比较/%
Table 1 Comparison of the Performance/%
模型 准确率 精确率 召回率 F2分数 BOVW 67.78 32.55 40.19 33.74 Mask R-CNN 86.91 65.91 61.72 62.70 本文方法 89.15 72.19 68.01 68.20 
下载: 导出CSV
-
[1] 李强, 张景发.不同特征融合的震后损毁建筑物识别研究[J].地震研究, 2016, 39(3):486-493 http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=dzyj201603018 Li Qiang, Zhang Jingfa. Research on Earthquake Damaged Building Extraction by Different Features Fusion[J].Journal of Seismological Research, 2016, 39(3):486-493 http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=dzyj201603018
[2] 陶和平, 刘斌涛, 刘淑珍, 等.遥感在重大自然灾害监测中的应用前景:以5·12汶川地震为例[J].山地学报, 2008, 26(3):276-279 http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=sdxb200803005 Tao Heping, Liu Bintao, Liu Shuzhen, et al. Natural Hazards Monitoring Using Remote Sensing: A Case Study of 5-12 Wenchuan Earthquake[J]. Journal of Mountain Science, 2008, 26(3):276-279 http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=sdxb200803005
[3] 眭海刚, 刘超贤, 黄立洪, 等.遥感技术在震后建筑物损毁检测中的应用[J].武汉大学学报·信息科学版, 2019, 44(7):1 008-1 019 doi: 10.13203/j.whugis20190070 Sui Haigang, Liu Chaoxian, Huang Lihong, et al. Application of Remote Sensing Technology in Earthquake-Induced Building Damage Detection[J]. Geomatics and Information Science of Wuhan University, ,2019, 44(7):1 008-1 019 doi: 10.13203/j.whugis20190070
[4] 贾宁, 赵鹏祥.浅谈众源地理数据[J].中国科技信息, 2013(12):42 http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=zgkjxx201312007 Jia Ning, Zhao Pengxiang. Crowdsourcing Geographic Data[J]. China Science and Technology Information, 2013 (12):42 http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=zgkjxx201312007
[5] 姜青香, 刘慧平.利用纹理分析方法提取TM图像信息[J].遥感学报, 2004, 8(5):458-464 http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=ygxb200405012 Jiang Qingxiang, Liu Huiping.Extracting TM Image Information Using Texture Analysis[J]. Journal of Remote Sensing, 2004, 8(5):458-464 http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=ygxb200405012
[6] 涂继辉, 眭海刚, 吕枘蓬, 等.基于基尼系数的倾斜航空影像中建筑物立面损毁检测[J].武汉大学学报-信息科学版, 2017, 42(12):1 744-1 748 http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=whchkjdxxb201712010 Tu Jihui, Sui Haigang, Lv Ruipeng, et al. Building Facade Damage Detection Based on the Gini Index from Oblique Aerial Images[J]. Geomatics and Information Science of Wuhan University, 2017, 42(12):1 744-1 748 http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=whchkjdxxb201712010
[7] Vetrivel A, Gerke M, Kerle N, et al. Identification of Structurally Damaged Areas in Airborne Oblique Images Using a Visual-Bag-of-Words Approach[J]. Remote Sensing, 2016, 8(3): 231-252 doi: 10.3390/rs8030231
[8] Kerle N, Fernandez G J, Gerke M.Urban Structural Damage Assessment with Oblique UAV Imagery, Object-Based Image Analysis and Semantic Reasoning[C].The 35th Asian Conference on Remote Sensing, Nay Pyi Taw, Myanmar, 2014
[9] Fernandez G J, Kerle N, Gerke M. UAV-Based Urban Structural Damage Assessment Using Object-Based Image Analysis and Semantic Reasoning[J]. Natural Hazards & Earth System Sciences, 2015, 15(6):1 087-1 101 http://www.tandfonline.com/servlet/linkout?suffix=CIT0014&dbid=16&doi=10.1080%2F10447318.2018.1427831&key=10.5194%2Fnhess-15-1087-2015
[10] He K, Gkioxari G, Dollár P, et al. Mask R-CNN[C]. The IEEE International Conference on Computer Vision, Venice, Italy, 2017
[11] Ren S, He K, Girshick R, et al. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6):1 137-1 149 doi: 10.1109/TPAMI.2016.2577031
[12] Long J, Shelhamer E, Darrell T. Fully Convolutional Networks for Semantic Segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 39(4): 640-651 http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=994776c264e86f91bd6bda7f694c7564
[13] Woo S, Park J, Lee J Y, et al. CBAM: Convolutional Block Attention Module[C]. The European Conference on Computer Vision, Munich, Germany, 2018
[14] He K, Zhang X, Ren S, et al. Deep Residual Learning for Image Recognition[C].IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, USA, 2016
[15] 谢飞, 穆昱, 管子玉, 等.基于具有空间注意力机制的Mask R-CNN的口腔白斑分割[J].西北大学学报(自然科学版), 2020, 50(1): 9-15 http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=xbdxxb202001002 Xie Fei, Mu Yu, Guan Ziyu, et al.Oral Leukoplakia (OLK) Segmentation Based on Mask R-CNN with Spatial Attention Mechanism[J].Journal of Northwest University(Natural Science Edition), 2020, 50(1):9-15 http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=xbdxxb202001002
-
期刊类型引用(28)
1. 吕峥，孙群，温伯威，张付兵，马京振. 顾及形状相似性的道路与居民地协同化简方法. 地球信息科学学报. 2024(05): 1270-1282 . 
百度学术
2. 铁占琦. 利用改进的Hausdorff距离匹配多尺度线要素. 地理空间信息. 2024(05): 62-65 . 百度学术
3. 王庆社，王雅欣，姜青香，郭思慧. “天地图·北京”多源道路数据融合关键技术研究. 北京测绘. 2024(06): 874-879 . 百度学术
4. 陈钉均，梁芮嘉. 基于特征聚类驾驶员服从度跟驰模型参数标定. 计算机仿真. 2024(10): 126-132 . 百度学术
5. 齐杰，王中辉，李驿言. 基于图卷积神经网络的道路网匹配. 测绘通报. 2023(12): 19-24+44 . 百度学术
6. 吴冰娇，王中辉，杨飞. 用于多尺度道路网匹配的语义相似性计算模型. 测绘科学. 2022(03): 166-173 . 百度学术
7. 蒋阳升，俞高赏，胡路，李衍. 基于聚类站点客流公共特征的轨道交通车站精细分类. 交通运输系统工程与信息. 2022(04): 106-112 . 百度学术
8. 周秀华，李乃强. 基于多种相似度特征的道路实体融合方法. 测绘通报. 2021(08): 102-105+157 . 百度学术
9. 秦育罗，宋伟东，张在岩，孙小荣. 顾及几何特征和拓扑连续性的道路网匹配方法. 测绘通报. 2021(08): 55-60 . 百度学术
10. 杨飞，王中辉. 河系几何相似性的层次度量方法. 地球信息科学学报. 2021(12): 2139-2150 . 百度学术
11. 程绵绵，孙群，季晓林，赵云鹏. 改进平均Fréchet距离法及在化简评价中的应用. 测绘科学. 2020(03): 170-177 . 百度学术
12. 赵元棣，田英杰，吴佳馨. 航空器飞行轨迹相似性度量及聚类分析. 中国科技论文. 2020(02): 249-254 . 百度学术
13. Wenyue GUO，Anzhu YU，Qun SUN，Shaomei LI，Qing XU，Bowei WEN，Yuanfu LI. A Multisource Contour Matching Method Considering the Similarity of Geometric Features. Journal of Geodesy and Geoinformation Science. 2020(03): 76-87 . 必应学术
14. 秦育罗，郭冰，孙小荣. 改进Hausdorff距离及其在多尺度道路网匹配中的应用. 测绘科学技术学报. 2020(03): 313-318 . 百度学术
15. 郝志伟，李成名，殷勇，武鹏达，吴伟. 一种基于Fréchet距离的断裂等高线内插算法. 测绘通报. 2019(01): 65-68+74 . 百度学术
16. 郭文月，刘海砚，孙群，余岸竹，丁梓越. 顾及几何特征相似性的多源等高线匹配方法. 测绘学报. 2019(05): 643-653 . 百度学术
17. 宗琴，彭荃，秦万英. 一种基于模糊矩阵的空间面对象相似性度量算法. 北京测绘. 2019(10): 1218-1221 . 百度学术
18. 李兆兴，翟京生，武芳. 线要素综合的形状相似性评价方法. 武汉大学学报(信息科学版). 2019(12): 1859-1864 . 百度学术
19. 周家新，陈建勇，单志超，陈长康. 航空磁探中潜艇目标的联合估计检测方法研究. 兵工学报. 2018(05): 833-840 . 百度学术
20. 郭宁宁，盛业华，吕海洋，黄宝群，张思阳. 径向基函数神经网络的路网自动匹配算法. 测绘科学. 2018(03): 45-50 . 百度学术
21. 张瀚，李静，吕品，徐永志，刘格林. 六角格网的弧线矢量数据量化拟合方法. 计算机辅助设计与图形学学报. 2018(04): 557-567 . 百度学术
22. 邵世维，刘辉，肖立霞，王恒. 一种基于Fréchet距离的复杂线状要素匹配方法. 武汉大学学报(信息科学版). 2018(04): 516-521 . 百度学术
23. 苏满佳，张逸鸿，谢荣臻，朱海飞，管贻生，毛世鑫. 连续软体机器人的结构范型与形态复现. 机器人. 2018(05): 640-647+672 . 百度学术
24. 宗琴，姜树辉，刘艳霞. 多尺度矢量地图中模糊相似变换及其度量模型. 测绘科学. 2018(11): 72-78 . 百度学术
25. 郭文月，刘海砚，孙群，余岸竹，季晓林. 利用最长公共子序列度量线要素相似性的方法. 测绘科学技术学报. 2018(05): 518-523 . 百度学术
26. 郭宁宁，盛业华，黄宝群，吕海洋，张思阳. 基于人工神经网络的多特征因子路网匹配算法. 地球信息科学学报. 2016(09): 1153-1159 . 百度学术
27. 杨亚辉. 利用几何相似性快速测量鱼重的数学模型. 电子技术与软件工程. 2016(20): 182-183 . 百度学术
28. 逯跃锋，张奎，刘硕，吴跃，赵硕，李强，冯晨. 一种基于斜率差和方位角的矢量数据匹配算法. 山东大学学报(工学版). 2016(06): 31-39 . 百度学术
其他类型引用(30)
计量
- 文章访问数: 2239
- HTML全文浏览量: 954
- PDF下载量: 358
- 被引次数: 58
