Ice thickness is recognized as a key factor in ensuring the safety of icebreakers. Estimating ice thickness from a video clip is very challenging as typical icebreaking scenes contain a lot of movements and interactions between ice patches in addition to the digital noises. This paper introduces a method that narrows an original icebreaking image to an area sufficiently small enough to show only the scene of turning ice patch that reveals the thickness. Reducing the image area helps us to focus on meaningful ice actions and thickness. The proposed approach can be usefully applied to any icebreaking video stream.
Interest in pioneering an Arctic sea route is increasing as the sea ice decreases due to global warming. An Arctic route would be rewarding in many ways, such as beneficial economic effects due to cruise distance reduction and arctic resource development, but it would be accompanied by huge risks due to icebergs. Ice thickness is a crucial factor when attempting to safely cruise through icy waters, and hence verifying and predicting the ice thickness in a route is required.
There are many ways to measure ice thickness. Utilizing cameras is effective in terms of time efficiency, cost, and safety. Recently, researchers have taken advantage of image processing techniques that identify ice thickness by analyzing the video clips of icebreaking (Kulovesi and Lehtiranta, 2014; Nam et al., 2013). The motivation behind using icebreaking images is based on the fact that the icebreaking action pushes the contacting boundary of ice down into the water and turns it towards the icebreakers, showing the section of ice patch. Generally, ice thickness can be estimated by eyeballing, but this method is limited, requiring an experienced and trained observer.
