Machine Learning Based Method for Detecting Tsunami Devastated Area Using TerraSAR-X Data

Abstract

A method for rapid detection of tsunami devastated areas using multi-temporal TerraSAR-X data is proposed. To develop the method, machine learning algorithm, a branch of artificial intelligence (AI), is applied. We focus on the multiple bounce reflection which is a specific feature on Synthetic Aperture Radar (SAR) data to estimate building devastated areas. Finally, classifiers which enable automated classifications of damage patterns into predicted damage classes were built. The evaluation of the model was conducted through cross-validation and the best accuracy was obtained as 89.2 %.