Building Damage Estimation After Natural Disaster Using Multi Satellite Source Data based on Machine Learning

Building damage assessment after natural disasters is an important task for disaster management. In order to provide detailed levels of post-event building damage, this research will develop a deep learning building classification model with an online database using transfer learning on both 2D and 3D data.

The damage level is categorized into no damage, minor damage, major damage and collapsed. Natural disasters, such as fires, earthquakes and tsunamis, can cause serious building damage. Hence, the theoretical contribution of this study is providing a developed method for building damage classification. The practical contribution is supporting the government and rescue teams in their disaster response and decision making. It can also provide quick information for disaster management.


Project Leader:
Professor Linlin Ge, University of New South Wales

PhD Student:
Chang Liu, University of New South Wales