欢迎访问《图学学报》 分享到:

图学学报 ›› 2022, Vol. 43 ›› Issue (4): 608-615.DOI: 10.11996/JG.j.2095-302X.2022040608

• 图像处理与计算机视觉 • 上一篇    下一篇

基于计算机视觉的建筑施工期临时结构损伤识别方法

  

  1. 1. 华南理工大学土木与交通学院,广东 广州 510640;
    2. 亚热带建筑科学国家重点实验室,广东 广州 510640
  • 出版日期:2022-08-31 发布日期:2022-08-15
  • 通讯作者: 邓逸川(1989),男,副研究员,博士。主要研究方向为建筑信息模型、计算机视觉
  • 作者简介:梁振宇(2000),男,本科生。主要研究方向为建筑信息模型、计算机视觉
  • 基金资助:
    广东省自然科学基金项目(2022A1515010174);广州市科技计划项目基础与应用基础研究项目(202201010338)

A computer vision based structural damage identification method for temporary structure during construction

  1. 1. School of Civil Engineering and Transportation, South China University of Technology, Guangzhou Guangdong 510640, China;
    2. State Key Laboratory of Subtropical Building Science, Guangzhou Guangdong 510640, China
  • Online:2022-08-31 Published:2022-08-15
  • Contact: DENG Yi-chuan (1989), associate researcher, Ph.D. His main research interests cover BIM, CV
  • About author:LIANG Zhen-yu (2000), undergraduate student. His main research interests cover BIM, CV
  • Supported by:
    Natural Science Foundation of Guangdong Province (2022A1515010174); Guangzhou Science and Technology Program (202201010338)

摘要:

建筑施工临时结构是施工现场的事故主要风险源。以往的基于振动的临时结构监测方法依赖于在预先分析确定的监测关键部位放置的加速度传感器。但由于临时结构存在构件搭设不规范、施工现场不确定性等因素,通过有限元分析等手段得到的监测关键部位可能与实际情况相差较大,存在不确定性。为此提出一种基于欧拉运动放大算法的临时结构损伤识别方法,充分利用计算机视觉技术的全域覆盖及监测高效的优点。采用数码摄像机采集临时结构的数字图像序列,经过基于相位的欧拉运动放大算法处理,获取运动放大后的数字图像序列;运用 Canny 边缘识别算法获取边缘图像序列并消除运动放大造成的噪声,通过基于形心的运动跟踪算法获取临时结构的位移时程数据,并利用快速傅里叶变换进行频谱分析;与预先建立的损伤动力指纹库进行对比判断临时结构的损伤状态。以存在 10 种损伤状态的门式脚手架为测试对象,证明该方法的可行性与适用性。与加速度传感器测量进行对比,该方法平均误差为 0.95%,满足临时结构损伤状态识别的精度要求。

关键词: 欧拉运动放大, 临时结构, 计算机视觉, 结构损伤识别, 损伤动力指纹

Abstract:

Temporary structure is the main risk source of construction site accidents. Previous vibration-based detection methods mainly focus on setting accelerometers on some pre-defined critical areas. However, due to the factors such as nonstandard component erection and uncertainty of the construction site for the temporary structure, the critical areas of the monitoring obtained from the analysis may vary dramatically from the reality. Therefore, this paper proposed a structural damage identification method for temporary structure based on phased-based Eulerian video magnification algorithm, making full use of the advantages of global coverage and efficient monitoring of computer vision technology. The digital image of temporary structure vibration collected by digital camera was firstly processed by phased-based Eulerian video magnification to acquire motion-magnified image sequence in the particular frequency bands. Then, the canny edge detector was employed to identify the edges in the image sequence and eliminate the noise resulting from the magnification. The edges in the image sequence were utilized to acquire time-history data of temporary structure displacement based on the geometry centroid, from which resonant frequencies could be obtained after Fourier transformation, and finally the damage states were identified based on the pre-established damage dynamic fingerprint
database. The applicability of the proposed method was discussed in the context of the frame scaffold experiments with 10 kinds of damage states. By comparing the results between camera measurement and accelerometer measurement, the proposed method can yield satisfactory performance with an average error of 0.95%, fulfilling the accuracy requirements of damage identification.

Key words: Eulerian video magnification, temporary structure, computer vision, structural damage identification; damage dynamic fingerprint

中图分类号: