•  
  •  
 

Abstract

In order to guarantee the safe navigation of vessels in icy regions and effectively capture sea ice risk information surrounding ships, this study introduces a sea ice identification and risk assessment process utilizing shipboard camera images. Firstly, based on the distribution characteristics of sea area, a hierarchical region segmentation algorithm is proposed to precisely remove interference factors and extract the region of interest. Then, a multivariate K-means clustering algorithm is proposed to classify sea ice classes based on various characteristics, including spatial similarity, sea ice area, sea ice thickness, gray scale, and other multivariate information. Finally, a method for small scale sea ice identification and risk assessment is constructed to calculate the navigation risk index and realize the visualization of risk early warning. The experimental results show that the small-scale sea ice identification and risk assessment system for polar navigation can well segment the interference such as the sea-sky line and ship’s hull, as well as achieve precise sea ice classification, enable reliable real-time sea ice risk visualization early warning, and ensure safety for polar navigation.

COinS