Special Session

Deep Learning & Computer Vision Aided Characterization of Geotechnical Processes

 

Session organizers

Prof. Ningjun Jiang, Southeast University

Xiaole Han, University of Hawaii at Manoa

Yijie Wang, The Hong Kong Polytechnic University

 

 Overview

As a traditional civil engineering subject, geotechnical engineering deals with geo-structures built on, in or with soils and rocks. The intrinsic uncertainty, complexity, and invisibility of such materials have made the recognition and understanding of geotechnical processes challenging. Traditional sensors like strain gauges, thermometers, and piezometers can only record one-dimensional (1D) data streams. Though geotechnical engineers can interpret the geotechnical process behind the data, the information obtained from 1D data could be limited in some cases. As s new round of new infrastructure stimulus is carried out in the United States and China, more state-of-the-art techniques shall be applied during the design, site investigation, construction process management, and maintenance stages. 

​

Recently, with the revolutionary progress in data collection and cyberinfrastructure, two-dimensional (2D) data like images and videos, even three-dimensional (3D) data like point clouds and mesh files, can be captured and stored. Since the data sizes have increased dramatically compared to 1D data, more robust tools are required to finish the analysis. More sophisticated methods involved with computer vision and deep learning technologies have been applied to digitize, visualize, and analyze such geotechnical problems. This special session will cover, but not limited to, the following topics:

 

• Digital Image Correlation (DIC) application in Geotechnical engineering

• Particle Image Velocimetry (PIV) application in Geotechnical engineering

• Photogrammetry and Lidar for 3D reconstruction of geo-structures and features extraction

• 3D geotechnical structure detection from point clouds

• Virtual Reality (VR) of 3D geotechnical and geological models 

• Deep-learning based image processing for geotechnical and geological features recognition, segmentation, and quantification

• Convolutional neural network (CNN) for sensor data extraction, identification, and processing