A team from Waterloo Engineering has won an international competition aimed at advancing the state-of-the-art in the mapping of sea ice.
The team included professors, researchers and graduate students in a remote sensing research group that is a subgroup of the larger Vision and Image Processing (VIP) Lab within the Department of Systems Design Engineering.
Winning team members (l-r) Muhammed Patel, Xinwei Chen, Javier Noa Turnes, Linlin Xu, Fernando Pena Cantu and Jinman Park of the remote sensing research group at the VIP Lab at Waterloo Engineering.
Sponsored by organizations including the Danish Meteorological Institute and the European Space Agency (ESA), the AutoICE competition challenged teams to build machine learning models for three sea ice parameters – sea ice concentration, stage-of-development and floe size – to enable automated mapping using data from an ESA satellite called Sentinel-1.
The Waterloo team developed an artificial intelligence (AI) model based on a multi-task deep convolutional neural network to retrieve the parameters from multi-sensor satellite data.
Advanced AI techniques such as spatial-temporal encoding, domain adaptation and multi-task learning were incorporated to boost the performance of the model.
“We are thrilled to win first place in this worldwide competition, which demonstrates our team's leading position in the field of sea ice remote sensing,” said Dr. Xinwei Chen, a postdoctoral researcher.
“It also motivates us to continue to have worldwide impact in future research concerning the combination of AI and remote sensing.”
30 teams submitted solutions
The winning Waterloo team received about $4,500 and an invitation to remotely present its solution to an expert team at an upcoming event. Academic publication is also in the works.
Thirty teams from numerous countries including Germany, Italy and Spain submitted solutions to the challenge.
The remote sensing research group, which is dedicated to the science of capturing data of the earth from planes or satellites, is headed by Dr. David Clausi and Dr. Andrea Scott, both systems design engineering professors.
Clausi and Scott were Faculty leads for the AutoICE team. Other key members were research professor Dr. Linlin Xu, doctoral students Jinman Park and Javier Noa Turnes, and master’s students Muhammed Patel and Fernando Pena Cantu.