Detecting Icebergs with Machine Learning: Research and Commercialisation

John Stott, Senior Lecturer in Astrophysics and Sonny Bailey, Senior Research Associate in Iceberg Detection and Commercialisation

Department of Physics, Lancaster University

Wednesday 22nd May, 1-2pm, Cottrell 2V1 and streamed on Teams (contact the seminar organiser for the Teams link).
This seminar is open to all staff, students and affiliates of the University of Stirling. The seminar is hosted by Biological and Environmental Sciences (BES). 

Who this might appeal to: This joint presentation will hopefully be of interest to people involved in Earth observation, particularly of the polar regions. It may also be of interest to people who work on computer vision techniques and machine learning/deep learning.

John Stott: Detecting icebergs with machine learning

Abstract: I will discuss iceberg detection in synthetic aperture radar (SAR) data using a convolutional neural network (CNN). The CNN was trained to identify icebergs using labelled data. This labelling was performed using freely available C-band SAR and optical imaging from the European Space Agency (ESA) Sentinel-1 and Sentinel-2 satellites respectively. We use the iDPolRAD filter technique (Marino et al. 2016) to enhance the icebergs in the SAR images. The project was inspired by our research which used a similar CNN technique to search for, and characterise, a certain class of astronomical object (galaxy clusters) in large images of the sky.

Bio: Dr John Stott is a Senior Lecturer in Astrophysics at Lancaster University, specialising in galaxy evolution and machine learning. He was a Physics undergraduate at the University of Oxford, then gained a PhD in Astrophysics from Durham University. This was followed by postdoctoral positions at Liverpool JMU, Durham and Oxford before finally gaining a permanent position at Lancaster.

Sonny Bailey: Commercialisation of an iceberg detector using machine learning

Abstract: Every year around 10,000 icebergs calve off the glacier tongues in Greenland. Climate change is resulting in more shipping lanes opening up and there is an estimated $10 billion in Arctic global trade annually. Iceberg collisions are of interest not only to the shipping market, but also the oil and gas, fishing, tourism/leisure, insurance and defence sectors. Here, the state of the current market interest in iceberg detection is summarised, the results gained from doing digital testing and engagement with these various stakeholders are outlined, followed by a discussion of how commercialisation may fit into the overall research scope of iceberg and sea ice detection using SAR.

Bio: Dr Sonny Bailey is a Senior Research Associate in Iceberg Detection and Commercialisation at Lancaster University. He was an Environmental Science graduate at the University of Portsmouth, followed by an Environmental Hazards master graduate. He gained his PhD in Environmental Science, specialising in iceberg detection using quad-polarimetric SAR at the University of Stirling before gaining his current post doc position at Lancaster.