Nov 25th, 2025, Tuesday. 3PM. Room TG23.
Speaker: François Chadebecq
Title: Optimal Transport for Sea Ice Dynamics: Challenges and Solutions in Motion Estimation
Abstract: In this seminar, I will present my current research investigation on utilizing Optimal Transport (OT) to model sea ice dynamics. The state of sea ice, including its drift, deformation, growth, and melt, is a sensitive indicator of the Earth’s climate, playing an important role in Ocean–Atmosphere Interactions and the refinement of climate models and forecasts.
While satellite remote sensing provides researchers with a rich, localized stream of data on polar regions, analysing this multi-modal data to accurately track the motion of sea ice features (such as ice floes, ridges, and leads) between successive images remains challenging. Changing ice appearance (including cracks and ridges), ambiguities in feature matching , and issues with data integration are notable bottlenecks.
After a brief review of current feature-tracking and motion estimation approaches, I will introduce the Optimal Transport problem. OT offers a rigorous mathematical framework that can yield physically consistent sea ice motion estimates and theoretically address the challenges associated with complex deformation. Using simple illustrative examples, I will highlight current bottlenecks in applying OT-based registration methods to sea ice data and open a discussion on potential solutions for overcoming these limitations.
Bio: I am a Lecturer in Computer Vision at the Middlesex University Department of Computer Science. Before joining Middlesex University I was a Research Fellow at University College London within the Computer Science Department and the Wellcome / EPSRC Centre for Interventional and Surgical Sciences (WEISS). My main research interests are geometric vision modelling and 3D reconstruction in computer vision. My efforts notably focused on applications in medical endoscopy, computer-assisted interventions and robotics.