Dynamic safety monitoring system for anchored vessel based on image recognition technology

Abstract

Against the backdrop of advancing the "Transportation Powerhouse" initiative, the innovative integration of next-generation information technology with traditional water transport disciplines has driven the rapid development of China’s water transport sector, while also imposing higher demands on vessel safety management. Addressing the characteristics of abnormal vessel behavior and actual monitoring needs across various scenarios in the Three Gorges navigation anchorage, this paper constructs a four-dimensional screening and evaluation system incorporating risk priority to quantitatively score and select research scenarios. Using a hybrid data collection method of “surveillance screenshots, AI generation, Photoshop processing,” we build an abnormal behavior image library and a software development dataset. By applying the You Only Look Once (YOLO) and Deep Simple Online and Real-time Tracking (DeepSORT) algorithms, we have developed a dynamic safety monitoring system for vessels awaiting lock passage, which accurately tracks dynamic anomalies such as vessel drifting due to anchor failure. This system enables targeted and efficient identification of abnormal vessel behavior across various scenarios, providing technical support for intelligent anchorage supervision.