Processing Collections of Geo-Referenced Images for Natural Disasters

Abstract

After disaster strikes, emergency response teams need to work fast. In this context, crowdsourcing has emerged as a powerful mechanism where volunteers can help to process different tasks such as processing complex images using labeling and classification techniques. In this work we propose to address the problem of how to efficiently process large volumes of georeferenced images using crowdsourcing in the context of high risk such as natural disasters. Research on citizen science and crowdsourcing indicates that volunteers should be able to contribute in a useful way with a limited time to a project, supported by the results of usability studies. We present the design of a platform for real-time processing of georeferenced images. In particular, we focus on the interaction between the crowdsourcing and the volunteers connected to a P2P network.