Building Damage Identification Method based on Mask R⁃CNN

YANG Jingsong; WANG Yuxin; LI Zhitao; LU Zewei; PENG Fumin

doi:10.13409/j.cnki.jdpme.20240304004

您当前的位置：

首页 >

文章列表页 >

Building Damage Identification Method based on Mask R⁃CNN

更新时间：2025-07-07

- Building Damage Identification Method based on Mask R⁃CNN
- Journal of Disaster Prevention and Mitigation Engineering Vol. 45, Issue 3, Pages: 562-570(2025)
- 作者机构：
  
  1.防灾科技学院资源与环境学院，河北三河 065201
  2.河北省地震灾害仪器与监测技术重点实验室，河北三河065201
  3.廊坊市精密主动震源重点实验室，河北三河 065201
  4.应急管理部国家自然灾害防治研究院，北京 100085
- 作者简介：
- 基金信息：
- DOI：10.13409/j.cnki.jdpme.20240304004
  CLC： TU375
- Received：04 March 2024，
  
  Revised：2024-08-05，
  
  Published：28 June 2025
- 稿件说明：
移动端阅览
杨敬松,王煜鑫,李智涛等.基于Mask R⁃CNN的多类建筑物损伤识别方法[J].防灾减灾工程学报,2025,45(03):562-570.

YANG Jingsong,WANG Yuxin,LI Zhitao,et al.Building Damage Identification Method based on Mask R⁃CNN[J].Journal of Disaster Prevention and Mitigation Engineering,2025,45(03):562-570.
杨敬松,王煜鑫,李智涛等.基于Mask R⁃CNN的多类建筑物损伤识别方法[J].防灾减灾工程学报,2025,45(03):562-570. DOI： 10.13409/j.cnki.jdpme.20240304004.

YANG Jingsong,WANG Yuxin,LI Zhitao,et al.Building Damage Identification Method based on Mask R⁃CNN[J].Journal of Disaster Prevention and Mitigation Engineering,2025,45(03):562-570. DOI： 10.13409/j.cnki.jdpme.20240304004.

摘要

地震发生后快速对建筑物损伤进行识别，可以提高灾害损失评估的效率，并为救援提供有效地决策支持。针对因背景干扰带来的重要特征表达能力弱的问题，提出一种基于深度学习框架Mask R⁃CNN的多建筑物损伤识别方法。首先，对样本图像进行预处理，克服复杂环境背景因素干扰，并进行多途径扩增，得到用于深度学习的扩增样本数据集。其次，优化特征提取网络，采用嵌入注意力机制模块SE的MobileNetv3网络作为主干网络，增加模型对建筑物损伤空间及语义信息的提取，有效避免背景对模型性能的影响，改进损失函数，避免遗漏类别和类别错分现象，同时引入迁移学习，降低训练成本；最后，采用定性分析和定量评估相结合的手段，多维度评估模型泛化能力和鲁棒性。改进后的Mask R‑CNN模型的平均精度达到了84.34%，相对于原始的Mask R‑CNN模型，精度提高了9.12%。结果表明，改进后的模型在识别含有多种损伤特征和噪声背景的建筑物损伤图像方面表现良好，可以为地震后建筑物的损伤评估提供有效地技术支持。

Abstract

After an earthquake

quickly identifying building damage can improve the efficiency of disaster loss assessment and provide effective decision-making support for rescue efforts. Background interference can weaken the expression of important features. To address this issue

a method based on the deep learning framework Mask R-CNN was proposed for identifying multiple types of building damage. First

the sample images underwent preprocessing to overcome interference from complex environmental backgrounds. Multiple augmentation techniques were applied to generate a dataset suitable for deep learning. Next

the feature extraction network was optimized using the MobileNetv3 network with an embedded SE (Squeeze-and-Excitation) attention mechanism module as the backbone. This design enhanced the model's ability to extract spatial and semantic information related to building damage and effectively reduced the negative impact of background interference. Furthermore

the loss function was improved to avoid category omissions and misclassifications

while transfer learning was introduced to reduce training costs. Finally

a combination of qualitative analysis and quantitative evaluation was employed to assess the model's generalization ability and robustness. The results demonstrated that the improved Mask R-CNN model achieved an average precision of 84.34%

which was a 9.12% improvement over the original Mask R-CNN model. The improved model performs well in identifying building damage images with various damage features and noisy backgrounds

providing effective technical support for post-earthquake building damage assessment.

关键词

Keywords

references

Goodfellow I ， Bengio Y ， Courville A . Deep learning ［M］. Cambridge ： MIT Press ， 2016 .

Koch C ， Georgieva K ， Kasireddy V ， et al . A review of computer vision-based defect detection and condition assessment of concrete and asphalt civil infrastructure ［J］. Advanced Engineering Informatic . 2015 ， 29 （ 2 ）： 196 - 210 .

Wu W ， Qurishee M A ， Owino J ， et al . Coupling deep learning and UAV for infrastructure condition assessment automation ［C］∥ Proceedings of the 2018 IEEE International Smart Cities Conference （ISC2） . Kansas City， MO， USA ：［s.n.］， 2018 .

Griffiths D ， Boehm J . Rapid object detection systems， utilising deep learning and unmanned aerial systems （UAS） for civil engineering applications ［J］. The International Archives of the Photogrammetry Remote Sensing and Spatial Information Sciences ，（ISPRS）， 2018 ， 42 ： 391 - 398 .

陈梦 . 基于深度学习的建筑物震害遥感识别研究［D］. 北京：中国地震局地震预测研究所， 2019 .

Chen M . Research on identification of building damage information from remote sensing image based on deep learning ［D］. Beijing ： Institute of Earthquake Forecasting， China Earthquake Administration ， 2019 . （in Chinese）

Lee Y I ， Kim B ， Cho S . Image-based spalling detection of concrete structures using deep learning ［J］. Journal of the Korea Concrete Institute ， 2018 ， 30 ： 91 - 99 .

陈鹏，汪本康，高飒，等 . 利用ResNet进行建筑物倒塌评估［J］. 武汉大学学报（信息科学版）， 2020 ， 45 （ 8 ）： 1179 - 1184 .

Chen P ， Wang B K ， Gao S ， et al . Building collapse assessment with ResNet ［J］. Geomatics and Information Science of Wuhan University ， 2020 ， 45 （ 8 ）： 1179 - 1184 . （in Chinese）

Hori M ， Sutoh A ， Saitoh Y . Strong motion measurement using security video camera ［J］. Journal of Structural Mechanics and Earthquake Engineering ， Japan Society of Civil Engineers， 2000 ， 647 （ 51 ）： 57 - 66 .

Arimitsu Yokota ， Takayuki Hamamoto ， Hisashi Koga . Estimation of earthquake ground motion by image analysis of sliding objects taken with a fixed camera ［C］∥ 21st International Conference on Pattern Recognition （ICPR 2012） . Tsukuba， Japan ：［s.n.］， 2012 .

Cha Y J ， Choi W . Deep learning⁃based crack damage detection using convolutional neural networks ［J］. Computer-Aided Civil and Infrastructure Engineering ， 2017 ， 32 （ 5 ）： 361 ⁃ 378 .

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］. Photogrammetry and Remote Sensing ， 2018 ， 140 ： 45 ⁃ 59 .

Yeum C M ， Dyke S J ， Ramire Z ， et al . Big visual data analysis for damage evaluation in civil engineering ［C］∥ The International Conference on Smart Infrastructure and Construction . Cambridge， UK ：［s.n.］， 2016 .

Gao Y Q ， Mosalam K M . Deep transfer learning for image‑based structural damage recognition ［J］. Computer⁃Aided Civil and Infrastructure Engineering ， 2018 ， 33 （ 9 ）： 748 ⁃ 768 .

Shelhamer E ， Long J ， Darrell T . Fully convolutional networks for semantic segmentation ［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence ， 2017 ， 39 （ 4 ）： 640 - 651 .

Hu J ， Shen L ， Albanie S ， et al . Squeeze-and-excitation networks ［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence ， 2020 ， 42 （ 8 ）： 2011 - 2023 .

Cai Z ， Vasconcelos C R . High quality object detection and instance segmentation ［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence ， 2021 ， 43 （ 5 ）： 1483 - 1498 .

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Lightweight Detection of Tunnel Water Leakage Based on Improved YOLOv8 Model

Spatiotemporal Forecast of Bluff‑body Wakes Using Artificial Intelligence Technologies

Related Author

TIAN Chenghang

ZHAO Limeng

LI Yonggang

ZHANG Fangjian

ZHANG Wengang

SUN Weixin

DING Xuanming

LIU Junle

Related Institution

School of Civil Engineering, Chongqing University

State Key Laboratory of Safety and Resilience of Civil Engineering in Mountain Area

Guangzhou Metro Construction Management Co., Ltd

China Railway 11th Bureau Group Co., Ltd

College of Aerospace Engineering, Chongqing University

⁰