Effectiveness of RGB Imagery from Diverse Sources for Real-Time Urban Flood Water Mapping

摘要

Timeliness is a critical requirement for the provision of information during disasters such as floods in urban areas. Images from RGB remote sensors (such as those carried on satellites, aircraft, UAVs and on the International Space Station) are potentially cost-effective data sources for real-time applications. This paper describes work undertaken to evaluate two high resolution RGB image datasets for rapid response mapping and monitoring of urban visible flood water extent. Our overall goal is to develop a robust methodology for universally applicable extraction of the flooded area information. The methods and results are demonstrated as a case study of the characterization and delineation of visible flood extent and its changes that occurred during the June 2013 city of Calgary flood event. The input imagery included very high resolution aerial photography and imagery acquired with the International Space Station's (ISS) SERVIR Environmental Research and Visualization System (ISERV). The methodology development involved analysis and comparison of the spectral responses of urban land surface types, shadows and turbid flood water. Based on an understanding of these spectral properties, a universally applicable method was developed and assessed to extract visible flood water from RGB imagery.