Reconnaissance of Golcuk 1999 Earthquake Damage Using Satellite Images

Abstract

This study utilizes remotely sensed pre- and post-disaster images in order to detect any change specifically associated with structural and major regional damage caused by natural disasters such as a strong earthquake. Currently, postdisaster reconnaissance is conducted using mostly ground-truth teams. Although by using this method high-resolution assessment can be provided, it is very time consuming, costly and sometimes not practical. This study investigates an unsupervised near near-time method of reconnaissance using a pair of coregistered remotely sensed images of the same scene acquired at different times as input. The output is an image (mostly binary) in which “changed” pixels are separated from “not-changed” ones. The approach is fully deterministic using principal component analysis, normalization and further setting a threshold to classify pixels. For verification, a set of multispectral images from Golcuk-Turkey 1999 earthquake was used. It is concluded the program generates very similar results as obtained by ground-truth teams.