Ship near-miss risk identification method based on multi-modal deep fusion model under complex weather and sea conditions

An effective risk identification method is crucial for improving ship navigation safety. However, traditional approaches often underestimate the importance of near-miss accidents, incidents that do not result in immediate damage but usually reveal hidden security vulnerabilities. To deal with this problem, this paper proposes a ship near-miss risk identification method based on multi-modal deep fusion model. First, this method fuses different data through multi-modal fusion technology. Secondly, a near-miss probability model was constructed through correlation analysis and kernel density estimation. Finally, machine learning algorithms are used to analyze high-dimensional data to assess potential weather-related risks faced by ships during navigation comprehensively. Experimental results show that this model can predict dangerous situations at an earlier stage with an accuracy rate of over 95 %, and effectively identify those ship behaviors that may not lead to immediate serious consequences but are high-risk. This method significantly improves the accuracy of shipping safety management, and provides more reliable and efficient decision support for ships. • This paper proposes a ship near-miss risk identification method based on multi-modal deep fusion model. • The method can deeply fusions multi-source data. • This method can enhance the risk model's ability to identify near miss accidents. • This method can provide more effective decision-making support for ships.