Integration of Building Unit Based and Zonal Based Approaches to Detect Washed-away Buildings in a Tsunami Affected Area Using TerraSAR-X Data

Abstract

A new method to detect buildings washed away by a tsunami is proposed using high-resolution Synthetic Aperture Radar (TerraSAR-X) data. The main focus of this study is to integrate building unit based approach and zonal based approach to detect building damage at a building unit scale. These were integrated based on decision tree application (C4.5) of machine learning algorithm. Finally, classifier for detecting washed-away buildings was developed and applied to Sendai coast, performing 91.0 % overall accuracy and kappa statistic of 0.82.