Vision-based Decision Support for Flood Risk Assessment Using Google Street View Images

Abstract

Flood is one of the most common and destructive natural disaster worldwide, both in terms of economic losses and human casualty. With the expected sea level rise for the next century, assessing flood risk become critical for coastal area residents and governments to make effective decisions about risk mitigation. To this end, Louisiana state in the U.S. have launched Comprehensive Master Plan and laid out $50 billion USD for coastal protection and restoration. However, most data for assessing flood risk in Louisiana are obsolete that were collected from street-level surveys in 1991 since it is too expensive to collect and update comprehensive data. As a result, this study proposes a vision-based approach using deep learning that can collect comprehensive data effectively and efficiently without human-involved street surveys. The proposed approach analyzes Google street view (GSV) images and simultaneously predicts several attributes of buildings that are necessary for assessing flood risks. The proposed approach can predict foundation type, height, and building type. Then, individual home or business owners can use a tool that visualize the flood risks to decide what measures to take to protect their assets. Also, local planners and policy makers will be able to develop “adaptation timelines” that prioritize infrastructure projects using the assessments of current and future flood risks.