Project Title: Discovering the way: Automated Machine Learning improvement of Ordnance Survey path network data
Host institutions: University of Leicester, Leicester, UK, co-hosted at Birkbeck, University of London and Ordnance Survey
Funding: Full scholarship for UK/EU students by Ordnance Survey
Application deadline: 9 August 2020
Start date: September 2020
Application link: https://le.ac.uk/study/research-degrees/funded-opportunities/gge-de-sabbata-os-2020
- Principal Supervisor: Dr Stefano De Sabbata (University of Leicester)
- Co-Supervisor: Dr Andrea Ballatore (Birkbeck, University of London)
- Co-Supervisor: Dr Stefano Cavazzi (Ordnance Survey)
Project: Geospatial Artificial Intelligence (GeoAI) is an emerging field (Li, 2020) that aims to combine methods and concepts in geographic information science with the transformative innovations in machine learning, particularly deep learning, and big data to create novel approaches to spatial data science (De Sabbata and Liu, 2019). The collection of geospatial data is a critical activity that faces many challenges, including high cost and variable data quality (Ballatore and Zipf, 2015). Current advances in GeoAI promise to enable more effective methods to detect and correct issues in complex real-world data.
Footpaths provide valuable opportunities for outdoor activities and exploration in UK, and in the urban environment are an important component of last-mile services. By investigating a novel spatially explicit machine learning technique that combines least-cost path network analysis this project aims to further improve the quality and completeness of OS MasterMap® Highways Network – Paths, the most accurate and authoritative path network dataset for Great Britain.The project will:
- extract a body of known footpaths from historic sources & current OS walking routes.
- develop a spatially explicit model to predict the location of footpaths based on a least-cost path (LCP) network analysis of topographic and environmental characteristics.
- apply a Deep Learning approach to identifying potential footpaths through graph link prediction.
- Compare predicted paths with the OS MasterMap® Highways Network – Paths data and establish a methodology to automatically update the OS path network.
This PhD offers the opportunity to work with two leading academic organisations (University of Leicester and Birkbeck) supported by the Ordnance Survey, Great Britain’s national mapping agency.
Keywords: geoAI, machine learning, geographic data science
Contact details: Stefano De Sabbata <email@example.com>