Buffer collision risk encounter identification based on incremental HDBSCAN clustering

Abstract

With the growing complexity of global maritime traffic, real-time monitoring of ship collision risks has become a critical task for ensuring maritime transportation safety. While existing research primarily focuses on collision risk assessment between ships, it still faces challenges in real-time identification of high-risk multi-ship encounter scenarios in complex waters. This paper proposed a buffer-based collision risk identification method using incremental HDBSCAN clustering, which aimed to monitor ship collision risks in real-time by introducing the concept of ship buffer zones and an incremental clustering mechanism. By modeling ships as dynamic moving areas and leveraging the HDBSCAN algorithm, the method adaptively identifies high-density and high-risk regions, while quantifying collision risk through the analytical calculation of DCPA, TCPA, and SDOI parameters. Experimental results reveal that the proposed method effectively captures high-risk ship clusters, demonstrates robustness in complex multi-ship encounter scenarios, and can be extended to support maritime surveillance and collision risk management. Therefore, this method offered a new technical approach for enhancing maritime traffic monitoring capabilities and optimizing risk management strategies.