All eyes on plankton: Applying cross-instrument data fusion in marine imaging

In this PhD project, you will advance the field of ecological imaging by integrating and analysing data from diverse imaging instruments. Marine ecosystems are studied using a variety of devices like the Underwater Vision Profiler (UVP), FlowCam, CPICS, ZooScan, and PlanktonImager. These instruments capture high-resolution data but vary significantly in scale, resolution, and differences in appearance due to types of imaging techniques, making cross-instrument analysis difficult. Additionally, the species distribution of most samples is long-tailed - a few organisms are very common, but many ecologically significant species are quite rare. This presents a challenging scenario for standard supervised AI models.

The project involves developing machine learning techniques to create adaptable data representations that harmonize and integrate heterogeneous datasets. You will investigate methods for combining data from different instruments at different stages of data analysis, aiming to enhance the classification of rare ecological features across multiple instruments. The research also includes designing an analysis framework capable of handling large, complex datasets and exploring the use of large pre-trained models for ecological imaging.

The project will provide extensive training in cutting-edge techniques, from data fusion to expandable machine learning, and includes opportunities to collaborate with leading marine and computer science research groups. The research is expected to significantly improve ecological monitoring and contribute to global efforts in environmental conservation and climate change impact assessment.

You will also be supervised by organisations other than the University of Southampton, including:

• Dr Eric Orenstein (lead supervisor), Mojtaba Masoudi and Dr Sari Giering from the National Oceanography Centre
• Dr Sophie Pitois from the Centre for Environment, Fisheries and Aquaculture Science