Automatic Damage Detection using Machine Learning and Deep Learning

The group is conducting research to analyze acceleration data of mobile phones and video images of cameras using machine learning and deep learning methods to detect cracks in pavements.

The acquired video image extracts features by the Linear Binary Pattern (LBP) method, and then classifies cracks by Support Vector Machine (SVM), one of the common machine learning methods. The raw data acquired by the accelerometer is reconstructed by Wavelet Transform (WT), and the reconstructed data are automatically classified by Convolution Neural Network (CNN), a method of deep learning. Acquiring accurate data is also necessary for monitoring techniques, but if a large number of data is analyzed using machine learning or deep learning methods, infrastructure damage can be more effectively and accurately identified.

Selected references

  • Chen, C., Seo, H. S., Zhao, Y., Chen, B., Kim, J. W., Choi, Y., & Bang, M. (2019). Automatic Pavement Crack Detection Based on Image Recognition. International Conference on Smart Infrastructure and Construction 2019 (ICSIC). ICE Publishing. doi:10.1680/icsic.64669.361
  • Chen, C., Seo, H. S., Zhao, Y., Chen, B., Kim, J. W., Choi, Y., & Bang, M. (2019). Pavement Damage Detection System Using Big Data Analysis of Multiple Sensor. International Conference on Smart Infrastructure and Construction 2019 (ICSIC). ICE Publishing. doi:10.1680/icsic.64669.559


 

Back to: School of Engineering