Research topic: Multisensor data fusion and applied machine learning for analysis of arctic sea ice.
Main Supervisor: Associate Professor Ekaterina Kim
Co-Supervisor: Professor Roger Skjetne
Background:
Research interests: Sensor fusion, computer vision, machine learning
The PhD task is part of a larger research project, DIGITAL SEAICE, which has the goal of building an infrastructure to monitor and predict ice conditions in the arctic. At the moment, satelite data is used for this purpose, but this has the drawbacks of yielding (relatively) low resolution measurements at large time intervals. Motivated by this, it is desired to integrate local observations from research vessels into the global estimation. Obtaining these local measurements is my main task.
The method for doing this is based on recent advances in sea ice segmentation on camera images using deep learning techniques. Using these segmented images, I will investigate how classical estimation techniques can be applied to retrieve different ice-parameters such as location, drift and thickness. Other sensors (radar/lidar) will also be investigated to yield a more robust estimator in times of degregaded visual conditions such as fog or at night time.