Detection of Subsurface Void from Radar Images by Three-dimensional Convolutional Neural Network and Finite Difference Time Domain Method

In this article, a deep learning algorithm to detect subsurface void from radar images by three-dimensional Convolutional Neural Network (3D-CNN) is developed. To obtain training data, producing 3D electromagnetic (EM) responses by three-dimensional finite difference time domain (3D-FDTD) method is...

Full description

Saved in:
Bibliographic Details
Published inSEISAN KENKYU Vol. 73; no. 5; pp. 327 - 331
Main Authors YAMAGUCHI, Takahiro, MIZUTANI, Tsukasa
Format Journal Article
LanguageEnglish
Published Institute of Industrial Science The University of Tokyo 01.11.2021
Online AccessGet full text

Cover

More Information
Summary:In this article, a deep learning algorithm to detect subsurface void from radar images by three-dimensional Convolutional Neural Network (3D-CNN) is developed. To obtain training data, producing 3D electromagnetic (EM) responses by three-dimensional finite difference time domain (3D-FDTD) method is not feasible, requiring large calculation cost. The methodology to reproduce 3D EM responses by two-dimensional finite difference time domain (2D-FDTD) method is proposed for the first time. The proposed method is validated by subsurface void measurement data. 3D-CNN trained by simulated data shows high performance indicating the effectiveness of 3D subsurface sensing.
ISSN:0037-105X
1881-2058
DOI:10.11188/seisankenkyu.73.327