Subsurface Pipe Detection by 3D Convolutional Neural Network and Kirchhoff Migration

• YAMAGUCHI Takahiro

  Institute of Industrial Science, The University of Tokyo, Japan

• MIZUTANI Tsukasa

  Institute of Industrial Science, The University of Tokyo, Japan

Bibliographic Information

Other Title

• 三次元畳込みニューラルネットワークとキルヒホッフマイグレーションによる地中埋設管の検知

Abstract

In this paper, the novel algorithm to detect subsurface utility pipes from Ground penetrating Radar (GPR) data is proposed. Due to the highspeed and dense 3D monitoring, GPR is a promising tool. However, vast amount of radar data and the difficulty of interpretation are the bottlenecks. We propose a new algorithm by the combination of 3D Convolutional Neural Network (3D-CNN) and Kirchhoff migration. A 3D-CNN model classifies data into transverse, longitudinal pipes and no pipe section. After detection by 3D-CNN, Kirchhoff migration is applied to extract peaks of section images as pipes’ 3D positions. Pipes are successfully visualized with reasonable calculation time.

Journal

• SEISAN KENKYU

  SEISAN KENKYU 72 (4), 329-333, 2020-07-01

  Institute of Industrial Science The University of Tokyo

Details

• CRID: 1390848250136539264
• NII Article ID: 130007888764
• NII Book ID: AN00127075
• DOI: 10.11188/seisankenkyu.72.329
• ISSN: 18812058
• ISSN: 0037105X
• NDL BIB ID: 030662351
• Web Site: http://id.ndl.go.jp/bib/030662351
• Web Site: https://ndlsearch.ndl.go.jp/books/R000000004-I030662351
• Text Lang: en
• Data Source:
  • JaLC
  • NDL
  • CiNii Articles
  • KAKEN
• Abstract License Flag: Disallowed