Author(s): S. Ansari; C. D. Rennie; L. Poirier

Linked Author(s):

Keywords:

Abstract: This research focuses on the development of a novel approach for monitoring and classifying different kinds of sea ice interacting with bridge piers in Northumberland Strait. Different ice types can have a varying impact on the navigability of a vessel or the loads on a structure. As such, the ability to monitor and classify different ice types automatically is an added strength to the existing ice load monitoring system present at the Confederation Bridge since 1997 and should allow for further automation. A deep learning algorithm based on Convolutional Neural Networks (CNN) has been utilized for training an algorithm to classify four different types of sea ice interacting with the bridge pier. Despite the fact that classification of different ice types from images is sometimes very challenging even when performed manually by experts, the developed algorithm is able to identify different classes accurately and on a real-time basis. The accuracy of the results demonstrates high practical utility of the method for similar applications.

DOI:

Year: 2020