A radar-based method for detecting tsunami devastated areas using machine learning algorithm

Hideomi Gokon, Joachim Post, Enrico Stein, Sandro Martinis, André Twele, Matthias Mück, Shunichi Koshimura

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

After catastrophic earthquakes and subsequent tsunamis, relief activity and reconstruction activity might be delayed due to the breakdown of information network and interception of roads to the devastated zones. To rapidly estimate the impact of the tsunami, air- or spaceborne remote sensing technologies can be used. In particular, Synthetic Aperture Radar (SAR) which is available independent of atmospheric conditions is promising. In this study, a semi-automatic method using high-resolution multi-temporal SAR data (TerraSAR-X) is proposed to estimate building damage in tsunami devastated areas related to the 2011 Tohoku earthquake tsunami. To develop the method, machine learning, a research field of artificial intelligence, is applied. Finally, evaluation of the model is conducted through cross-validation. The best accuracy is obtained as 89.2 % and kappa statistic is calculated as 0.76 when a decision tree approach (C4.5) is applied.

Original languageEnglish
Title of host publication34th Asian Conference on Remote Sensing 2013, ACRS 2013
PublisherAsian Association on Remote Sensing
Pages3535-3541
Number of pages7
ISBN (Print)9781629939100
Publication statusPublished - 2013 Jan 1
Event34th Asian Conference on Remote Sensing 2013, ACRS 2013 - Bali, Indonesia
Duration: 2013 Oct 202013 Oct 24

Publication series

Name34th Asian Conference on Remote Sensing 2013, ACRS 2013
Volume4

Other

Other34th Asian Conference on Remote Sensing 2013, ACRS 2013
CountryIndonesia
CityBali
Period13/10/2013/10/24

Keywords

  • Building damage
  • Change detection
  • Machine learning
  • TerraSAR-X
  • Tsunami

ASJC Scopus subject areas

  • Computer Networks and Communications

Fingerprint Dive into the research topics of 'A radar-based method for detecting tsunami devastated areas using machine learning algorithm'. Together they form a unique fingerprint.

Cite this