The Marine Technology Research Unit is dedicated to advancing methods for visualizing and analysing marine environments through state-of-the-art 3D imaging and analysis. Our research integrates advanced digital and machine-learning techniques with marine science to address critical environmental challenges. This commitment aligns with the strategic objectives of UK Research and Innovation (UKRI), the Natural Environment Research Council (NERC), and the Advanced Research and Invention Agency (ARIA), promoting innovative approaches to understanding and safeguarding our oceans.
We use orthomosaics—high-resolution, large-scale, geo-referenced images composed of multiple photographs stitched together—to produce accurate maps of the seafloor and coastal areas. These mosaics allow for consistent monitoring and detailed analysis of marine ecosystems, supporting long-term assessments of environmental resilience. Through photogrammetry, we derive precise, measurable data from overlapping photographs, creating 3D models that enable comprehensive spatial analyses essential for habitat assessment, conservation planning, and environmental impact monitoring.
Our team employs Structure from Motion (SfM) and Simultaneous Localization and Mapping (SLAM) techniques to reconstruct 3D structures from 2D images and videos. SfM has become a cornerstone for marine 3D imaging due to its flexibility and scalability, enabling high-resolution models from simple photographic data. Visual SLAM augments these models by integrating real-time data from multiple cameras or other sensors, creating seamless, real-time 3D maps.
We also integrate multi-camera sensing setups, which involve multiple synchronized cameras to capture comprehensive perspectives of marine habitats. This approach allows us to rapidly generate accurate 3D representations and assess structural complexity, facilitating detailed studies of biodiversity and ecological dynamics.
Through semantic labelling, we add value to the raw 3D models by identifying and categorizing key features within marine habitats, such as coral species, rock formations, and man-made structures. These labels facilitate automated monitoring of biodiversity and habitat conditions, providing essential data for conservation and restoration. Deep learning models are integral to our analytical workflow, particularly in the classification and identification of marine species and habitats. Our team applies convolutional neural networks (CNNs) to analyse visual and audio data, enhancing our ability to detect and monitor species in challenging underwater conditions.
The data and models produced by our 3D imaging and analysis efforts contribute directly to conservation policy and management. Our high-resolution visual data provides stakeholders with actionable insights into ecosystem health, enabling informed decision-making for marine protected areas, resource management, and climate resilience planning.
MTRU is committed to developing and transferring digital skills essential for marine research. Through hands-on training and knowledge transfer collaborations, we equip researchers and practitioners with the skills needed to harness emerging technologies like photogrammetry and AI-driven analysis. This supports a new generation of environmental scientists skilled in digital methodologies for sustainable ocean stewardship.