欢迎访问《图学学报》 分享到：
图学学报

图学学报 ›› 2022, Vol. 43 ›› Issue (4): 721-728.DOI: 10.11996/JG.j.2095-302X.2022040721

• 建筑与城市信息模型 • 上一篇    下一篇

基于计算机视觉与 BIM 的裂缝可视化管理方法

  

  1. 1. 深圳大学中澳 BIM 与智慧建造联合研究中心，广东 深圳 518060；
    2. 深圳大学土木与交通工程学院，广东 深圳 518060；
    3. 北京科技大学土木与资源工程学院，北京 100083
  • 出版日期:2022-08-31 发布日期:2022-08-15
  • 通讯作者: 许镇(1986)，男，教授，博士。主要研究方向为城市数字防灾
  • 作者简介:熊琛(1990)，男，副教授，博士。主要研究方向为地震工程与城市防灾
  • 基金资助:
    广东大学生科技创新培育专项资金项目(pdjh2020b0505)；国家重点研发计划课题(2021YFF0501002)；北京市自然科学基金面上项目
    (8212011)

Crack visualization management method based on computer vision and BIM

  1. 1. Sino-Australia Joint Research Center in BIM and Smart Construction, Shenzhen University, Shenzhen Guangdong 518060, China;
    2. College of Civil and Transportation Engineering, Shenzhen University, Shenzhen Guangdong 518060, China;
    3. School of Civil and Resource Engineering, University of Science and Technology Beijing, Beijing 100083, China
  • Online:2022-08-31 Published:2022-08-15
  • Contact: XU Zhen (1986), professor, Ph.D. His main research interest covers urban digital disaster prevention
  • About author:XIONG Chen (1990), associate professor, Ph.D. His main research interests cover earthquake engineering and urban disaster reduction
  • Supported by:
    Guangdong University Student Science and Technology Innovation Cultivation Special Fund Project (pdjh2020b0505); National Key
    R&D Program of China (2021YFF0501002); Beijing Municipal Natural Science Foundation (8212011)

PDF (PC)

177

摘要：

对结构表面裂缝进行持续检监测与管理对保障结构安全具有重要意义。为实现结构裂缝自动识别与管理，提出了一种基于计算机视觉与建筑信息模型(BIM)的裂缝识别、矢量化处理与可视化管理方法。首先基于深度学习的图像识别方法，从相机拍摄的结构表面图像中提取出裂缝形态的栅格图像；其次，提出了一种裂缝形态栅格图像的自动矢量化方法，获取裂缝形态关键点坐标；最终，使用 Dynamo 程序实现裂缝 BIM模型的自动建模与可视化。该方法可以获取裂缝的拓扑形态信息，并显著降低裂缝信息的存储数据量与可视化难度。通过开展 BIM 构件的碰撞分析，还可准确识别裂缝属于结构中的哪个构件，将裂缝所属的构件编号信息与裂缝宽度信息作为裂缝图元参数写入 BIM 模型，实现裂缝矢量化与裂缝 BIM 模型自动化建模与管理，为大范围、大批量的裂缝自动化检监测与管理提供参考。

关键词: 裂缝识别, 建筑信息模型, 裂缝可视化, 矢量化, 计算机视觉

Abstract:

Continuous monitoring and management of structural surface cracks is important to structural safety. To achieve automated structural crack identification and management, a series of crack identification, vectorization, and visualization methods were proposed based on computer vision and building information modeling (BIM). Firstly, the raster images of crack skeleton were extracted from structure surface images based on a deep learning method. Secondly, an automated vectorization method for the raster images of crack skeleton was proposed to obtain the coordinates of key points of cracks. Finally, the automated modeling and visualization of cracks were realized using Dynamo programming on BIM platform. The proposed crack vectorization method can obtain the topological information of cracks and significantly reduce the amount of stored data, thus facilitating crack visualization. In addition, through the collision analysis of BIM components, to which components the cracks belonged to can be easily identified. The component information and the crack width information can be stored as attribute data of each crack. The proposed method can attain an automated crack vectorization and visualization, providing a useful reference for large-scale crack identification and management.

Key words: crack identification, building information modeling, crack visualization, vectorization, computer vision

中图分类号: