Analysis of Earthquake-Caused Damages Using Multi-temporal Synthetic Aperture Radar (SAR) Imagery for Rapid Mapping

Abstract

Automated Earthquake Damage analysis is rapidly becoming a field of interest for researchers due to the huge research gaps and potentials it holds for developments. In addition, the significant losses of life and infrastructure has greatly compelled researchers to work in this domain. Since the latest developments for various disaster responses efforts are taking place using high resolution satellites, these real time images are also used for damage analysis of earthquakes. Satellites have the ability to provide us multi temporal data which is obtained in real time. However, the use of post event imagery obtained from satellites still holds a lot of challenges since there is a large scale shortage of training data that is available for this purpose, making the development of resilient algorithms a challenge. Bearing in mind the challenges and gaps of the previous studies, the current research will make use of SAR as well as Optical based satellite imagery for damage analysis. The performance is then evaluated by comparison with baseline methods and evaluation metrics to show that this research has out-performed the existing researches. The primary motive behind this study will be to identify the damaged buildings in real time from the data acquired from satellites. The dataset in this research makes use of SAR data that is compiled from Capella space dataset while the optical data comes from Maxar’s open space program. We consider this task as an image classification task, where we have two classes of damaged and intact. Machine learning algorithms will be applied on the compiled data so that the efficiency of damage detection of the overall system can be assessed. In this way a damage analysis system will be proposed which can help scientists detect earthquake damages in real time allowing for faster recovery efforts.