Early warning threshold analysis of maritime accidents based on image segmentation technology

Maritime channels are transportation carriers for national and regional trade. Accurately identifying the threshold for early warning systems for maritime emergencies can improve the accuracy of risk warnings and reduce the occurrence and harm resulting from maritime accidents, and thus improve the safety and security of maritime channels. Here, image segmentation technology is used to establish an early warning threshold selection model, and evaluation indicators such as the Youden index and five-fold cross-validation are used to verify the accuracy of the model. Empirical analysis based on past maritime accidents shows that the proposed model can accurately assess the risk of maritime accidents. The maximum entropy method has the highest warning accuracy, and the p-tile parameter method has the highest non-warning accuracy, which is used to minimise the false-alarm rate. The research results have important reference significance for identification of risk early warning thresholds and the construction of risk warning systems.