The Multi-Parameter Monitoring Method of Sea Ice Based on Image Processing Technique

Periodic sea ice is generated every winter seriously affects the safety of navigation and maritime operations in the high latitude sea areas. However, the current monitoring techniques could not detect the basic sea ice physical parameters in an automatic and accurate way, which is far away from the needs of marine production and navigation. At present, the most used sea ice monitoring methods based on naked eyes and equipment are time-consuming and labor-expensive. While the methods utilizing satellite remote sensing and radar observation station are highly limited by the accuracy and cost. In this paper, a method integrating image processing and machine learning technologies is proposed to automatically obtain the multi-parameters of sea ice, which include sea ice thickness, density, size distribution, and types. Our experiment results show that the average measurement error of thickness was less than 4 cm, the intensity was 2.8%, the size distribution was 5.8%, and the accuracy rate of type classification was 95%. This method has the advantage of easy operation, high precision, and low cost, which can greatly reduce the risk of manual operation and observation errors. It is expected to become the main method of sea ice monitoring.