Infrared thermography and 3D-data fusion for architectural heritage
Publication Date
July 27, 2026
Creator
Sutherland, Neil
University of Nottingham
Abstract
Comprehensive documentation is the foundation of effective conservation, repair and maintenance (CRM) practices within architectural heritage. In order to diagnose historic buildings and inform decision making, a combination of multidisciplinary surveys is fundamental to understanding a building’s significance and performance. InfraRed Thermography (IRT), a non-contact, non-invasive and non-destructive imaging technique, represents an established tool for heritage conservators, allowing both qualitative and quantitative assessments of temperature to be undertaken. However, the inherent low spatial resolution of thermal imaging has led recent work to fuse thermographic and geometric data in order to generate accurate 3D representations of architectural heritage encapsulating temperature information. This thesis presents InfraRed Thermography 3D-Data Fusion (IRT-3DDF), an emerging field of research transforming traditional two-dimensional thermal infrared images into novel three-dimensional models capable of visualising thermal anomalies. To explore the use of 3D thermal models for architectural heritage, IRT-3DDF is conceptualised as a discipline, expressing the key considerations required for effective thermal data capture and fundamental procedures required for the use of thermal infrared (TIR) cameras both geometrically and radiometrically.
The originality of the presented work is encapsulated by the direct matching of infrared and visible spectrum images for data fusion processes, a previously under-researched area of IRT-3DDF due to the difficulty determining correspondences across image modalities. This thesis introduces deep learning-based image matching, pre-trained neural networks designed for multi-view feature detection and feature matching, as a means of co-orienting multi-modal images depicting architectural heritage. Importantly, these correspondences are determined using ‘out-of the-box’ neural networks pre-trained on vast quantities of curated RGB images, utilised outside of their expected training domain when applied to TIR images and IRT-3DDF. The outcomes of this research demonstrate that deep learning-based image matching, when used in conjunction with combined bundle block adjustments, provide a viable approach for IRT-3DDF, capable of co-registering blocks of multi-modal images across varying scales, settings, sensors and subjects. Notably, the presented methods are fully-automatic, obviating the need for: geometric pre-calibration; sensors in fixed relative orientation; manual co-registration or alignment; associated position or orientation information; or simultaneous data capture. This thesis presents an efficient, cost-effective and flexible method of IRT-3DDF that can be readily-applied for multi-modal, multi-temporal and multi-sensor fusion.
Item Type
ethesis
Thesis Type
PhD
Supervisors
Marsh, Stuart
University of Nottingham
Mills, Jon
Newcastle University
Priestnall, Gary
University of Nottingham
Paul Bryan
Fabio Remondino
Fondazione Bruno Kessler
Subjects (LC)
Associated Schools / Departments
Related Publication URL(s)
UoN Repository URI
Except where otherwise noted, this item's license is described as
File(s)![Thumbnail Image]()
Name
Sutherland et al. - InfraRed Thermography and 3D-Data Fusion for Architectural Heritage
Description
InfraRed Thermography and 3D-Data Fusion for Architectural Heritage
Size
116.72 MB
Format
Adobe PDF
Checksum (MD5)
e91a8557a0f6f90a57f8ebd4b1fcd327