Development of Retaining Wall Condition Measurement Technology Using Deep Learning Models
Mr. Song Chang-ho (Ph.D. candidate) from the Department of Ocean Engineering at Pukyong National University received the Excellent Paper Award at the 2024 Fall Conference of the Korean Geotechnical New Materials Society.
At the recent conference held at the Future Hall of Pukyong National University, Mr. Song was awarded for his paper titled "Development of Retaining Wall Condition Measurement Technology Using CNN and Stereo Vision Cameras" (supervised by Professor Kim Yoon-tae).
Song Chang-ho, Ph.D. candidate, Department of Ocean Engineering, Pukyong National University. Provided by Pukyong National University
In this paper, he developed a technology that monitors the displacement of retaining walls using two stereo cameras and CNN, enabling real-time evaluation of the retaining wall’s displacement and settlement, which received high praise.
He proposed a technique in the paper that, through a deep learning model, extracts natural features of the retaining wall and calculates displacement without artificial targets. To automatically extract and identify retaining wall blocks from camera images using stereo cameras, four CNN-based image segmentation models?U-Net, FPN (Feature Pyramid Network), PSPNet (Pyramid Scene Parsing Network), and LinkNet?were applied to measure the condition of the retaining wall.
Since its establishment in 2001, the Korean Geotechnical New Materials Society has contributed to the development and advancement of geotechnical new materials through active academic activities in manufacturing, design, and construction fields related to geotechnical engineering and functional new materials for construction.
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
