Loading...

Machine Learning-Based Rapid Visual Screening of Buildings for Potential Seismic Damage Using BuildingImages

Shourabi, Shayan | 2022

840 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 55433 (09)
  4. University: Sharif University of Technology
  5. Department: Civil Engineering
  6. Advisor(s): Bakhshi, Ali
  7. Abstract:
  8. Rapid visual screening is a way to assess unsafe structures and reduce urban earthquake vulnerability. Although this method is faster than detailed analytical methods(e.g. Incremental dynamic analysis), it still requires a lot of time and financial resources to execute. Due to the widespread use of computer vision in engineering, this study has used this technology to perform rapid visual screening, improve its overall process and provide a standard method for automatic assessment of structures. In the process of rapid visual screening, the screener identifies features from the exterior view of the building, which are finally recorded in a scoring system. According to the score obtained, further decisions for seismic retrofitting can be made. In this thesis, first, the features that can be extracted by computer vision have been identified; Then, according to extract each feature, a database is created. After that, we benefit from convolutional neural networks such as ResNet, VGG, Inception, and DenseNet to classify those features. Eventually based on the final classifiers a computer program is introduced to help rapid visual screening process. Based on the fact that all previous research couldn't be used directly in rapid visual screening, the innovation of this project is to present a standard method based on computer vision and create required image databases, to use in the rapid visual screening process. The final algorithm achieved 61% to 96% accuracy in extracting different features for rapid visual screening. Also, in the case of soft-story recognition, this study increased the accuracy to 85% which shows a 0.62% improvement over the latest studies.
  9. Keywords:
  10. Computer Vision ; Convolutional Neural Network ; Seismic Assessment ; Seismic Damage ; Rapid Damage Detection ; Probabilistic Seismic Structures Vulnerability ; Seismic Vulnerability ; Machine Learning

 Digital Object List

 Bookmark

...see more