About the project
Photography is a key tool in monitoring remote marine habitats, but determining real ecological change is hampered by a lack of comparability between rapidly improving camera systems. This project will maximise comparability in image-derived ecological metrics by evaluating and minimizing sources of method bias, considering human- and AI-generated data.
Underwater photography is a critical component of characterization and repeated monitoring of remote marine environments, photography datasets are able to measure rapid environmental change. Advances in camera, lighting and battery technologies have improved image quality and the cost efficiency of these systems, while their use on autonomous vehicles have increased seafloor survey areas. However, changing photographic systems between monitoring points presents a challenge for generating comparable data that is critical for understanding real change in seafloor communities, as the ability to detect and identify organisms in imagery is highly influenced by image quality.
The speed of technological development means that photographic systems become obsolete and/or are updated frequently during a monitoring time series. A compounding factor is the use of artificial intelligence to generate biological data from images; while this can increase the volume of biological data, it may also magnify bias in source data. A standardized protocol for maximising comparability and reducing method bias that is generalizable to any camera system is needed.
The aims of this project are to design such a protocol:
- assess magnitudes of method bias in biological data from across multiple camera systems
- develop a generalized protocol for comparing derived ecological metrics across differing camera systems (including human- and AI-generated data)
- rank factors in terms of influence on data consistency for evaluation when selecting future camera systems.
The project will draw on existing photographic datasets captured with several camera systems on towed camera platforms, autonomous vehicles and remotely-operated vehicles.
You will also be supervised by organisations other than the University of Southampton, including Dr Daniel Roper and Dr Jennifer Durden from the National Oceanography Centre.
