Crowdsourced Map Building for Disaster Response
PI: Steven Reece
Department: Engineering Science
This project enables rapid delivery of information to disaster responders to improve rescue of people and recovery, addressing three gaps in data translation, identified from real-world experience, combining machine learning and crowd-sourcing technologies within their established Planetary Response Network (PRN). We will work with the UK NGO disaster risk reduction and response organisation, Rescue Global, to co-design a situation awareness tool to map post-disaster damage using satellite imagery. The team have worked alongside Rescue Global most recently in 2017 following Hurricanes Irma and Maria. The proposed project benefits thousands of victims following natural disasters by helping responders plan and efficiently execute relief operations by addressing the twin problems of data overload and lack of actionable intelligence immediately following a natural disaster.
The three gaps in the current system require:
- all laborious manual data processing steps to be automated to ensure rapid response to disaster events.
- a new satellite data processing pipeline to re-establish and retain data links to satellite imagery providers.
- the ability to generate immediately useful data products for disaster relief teams on the ground.
The project will improve PRN's data processing pipeline including how the system obtains, processes, uploads and extracts data to/from the Zooniverse platform and how the system generates maps for use by disaster responders. If successful we will be able to deploy PRN more frequently and with a wide range disaster response organisations by responding to requests for help we have, so far, been unable to provide. Furthermore, success will help in other (commercial) sectors, such as in our work with partners to classify land-use in Africa and monitor water resources in Afghanistan under an Oxford University Innovation consultancy contract.