Geospatial Systems - Thomas Goldring
In partnership with
Profile
Thomas Goldring
Biography
About Tom
Before joining the CDT in 2023, I worked as a Geo-environmental Engineer and Geospatial Specialist at Geoinvestigate Ltd, where I supported the development of mine shaft risk mapping and UAV-assisted slope stability assessment. I hold an integrated Masters in Geology from the University of Leicester, where my thesis used 3D modelling to reinterpret the fluvial architecture of the Wealden Formation.
My current research focuses on deep hierarchical learning for the classification and segmentation of national-scale point clouds. This work explores how National Mapping and Cadastral Agencies (NMCAs) can apply AI-driven and scalable approaches to maintain and update 3D geospatial data. Key themes include multitask and hierarchical learning for contextual 3D classification, 2D–3D knowledge distillation from Ordnance Survey MasterMap to point cloud data, and automatic error detection in large-scale building footprint datasets. The project contributes toward building a framework capable of national-scale 3D data fusion, validation, and maintenance, supporting the transition of national mapping into the era of digital twins.
Research
My research addresses a range of challenges associated with automated maintenance of big data. The project aims to develop a multi-branch pipeline that utilises contextual information from 2D data and new 3D data to update and generate mapping information. This will leverage the growing research in deep learning and handling big data.
We will use new photogrammetry-derived point clouds, of varying density, focused on the Isle of Wight, in combination with legacy mapping data from the Ordnance Survey to generate a deep learning model for semantic segmentation. This will provide a foundation to offer new, accurate information for the Ordnance Survey to use.
This project will support the Ordnance Survey in efficiently managing and updating 3D data by introducing scalable methods and integrating historical and model-based geospatial data.
