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

Authors: Yuan Ji; Jing Lu; Meizhi Jiang; Wan Su

Addresses: College of Transportation Engineering, Dalian Maritime University, Dalian, China ' College of Transportation Engineering, Dalian Maritime University, Dalian, China ' Transport Development Research Centre, Zhejiang Scientific Research Institute of Transport, Hangzhou, China ' College of Transportation Engineering, Dalian Maritime University, Dalian, China

Abstract: 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.

Keywords: early warning threshold; threshold selection model; Bayesian network; maritime accidents.

DOI: 10.1504/IJSTL.2024.143139

International Journal of Shipping and Transport Logistics, 2024 Vol.19 No.2/3, pp.278 - 296

Received: 07 Mar 2023
Accepted: 26 May 2023
Published online: 04 Dec 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article