Scientific Background
Marine litter threatens ocean health and the coastal livelihoods that depend on it. Scalable automated methods to detect, classify and quantify litter are needed to improve understanding of its sources, pathways and impacts, yet reliable monitoring at scale remains challenging. This PhD will help address that gap, supporting marine litter assessment in the United Kingdom, Europe and beyond, contributing evidence relevant to international action on plastic pollution.
The Centre for Environment, Fisheries and Aquaculture Science (CEFAS) is developing a visible-light deep learning classifier trained on a dataset covering 89 categories of litter, alongside laboratory facilities for characterising material reflectance. This project will extend that work using multispectral and hyperspectral imaging to improve discrimination between material types. A key challenge is to develop robust methods that account for sensor sensitivity, changing daylight illumination and diverse material reflectance properties.
Research methodology
You will combine existing labelled visible-light images with spectral measurements collected using laboratory imaging systems spanning approximately 380 to 1650 nanometres. You will characterise representative litter materials, curate imaging datasets, and develop machine-learning models for material detection and classification.
You will investigate model performance under changing illumination and across different imaging sensors. Approaches will include device-independent representations, transfer learning and domain adaptation. You will evaluate model robustness and explore translation into practical multispectral systems for large-scale drone-based monitoring.
Training
You will be based in the Colour & Imaging Laboratory within the School of Computing Sciences and work closely with scientists and engineers at CEFAS. Training will include imaging principles, laboratory measurement, computer vision, deep learning, geospatial analysis and experimental design, with potential opportunities for fieldwork including drone-flying. You will develop expertise in spectral imaging, artificial intelligence, environmental monitoring, data analysis and interdisciplinary research.
Person specification
We seek an enthusiastic and curious individual with a scientific, quantitative or computational background and a genuine interest in environmental challenges. Applicants from computer science, mathematics, physics, engineering, environmental science or biological science are warmly encouraged. Prior expertise across all these areas is not expected. Curiosity, analytical thinking and willingness to learn are essential.
Funding details
ARIES studentships are subject to UKRI terms and conditions(opens in a new window). Successful candidates who meet UKRI’s eligibility criteria will be awarded a fully-funded studentship, which covers fees, maintenance stipend (£21,805 p.a. for 2026/27) and a research training and support grant (RTSG). A limited number of studentships are available for international applicants, with the difference between 'home' and 'international' fees being waived by the registering university. Please note, however, that ARIES funding does not cover additional costs associated with relocation to, and living in, the UK, such as visa costs or the health surcharge.
