Facing a complex and ever-changing international environment, the maritime and shipping industry requires more efficient and precise data collection and analysis technologies to enhance management efficiency. A research team at The Maritime Data and Sustainable Development Centre (PMDC) at The Hong Kong Polytechnic University (PolyU) has developed a series of innovative artificial intelligence (AI) and big data driven tools, including advanced technology for estimating the supply and demand for typhoon shelter berths in Hong Kong to improve vessel monitoring and emergency response, a shipping data analytics platform that uses the Automatic Identification System (AIS) to assess real-time port congestion index and other maritime statistical indicators, and a trajectory analysis technology to effectively detect illegal fishing activities. These innovations not only provide governments and industry stakeholders with cutting-edge management solutions but also drive the sector's digital transformation and sustainable development.
Automatically identifying vessels and estimating the supply and demand for typhoon shelter berths
As a coastal city frequently affected by typhoons, Hong Kong must take effective preventive measures to ensure the safety of vessels. The research team, led by Prof. YANG Dong, Associate Head and Associate Professor of the PolyU Department of Logistics and Maritime Studies and Director of the PMDC, has collaborated with the Hong Kong Marine Department to develop an innovative monitoring technology. It utilises images of local vessels captured by unmanned aerial vehicles (UAVs), combined with deep learning-based computer vision algorithms to automatically identify and classify ships, achieving an accuracy rate of 98.6%. This new technology is used to predict the supply and demand for local vessel typhoon shelter berths from 2022 to 2035, optimising the design of berth management plans. The method significantly enhances the government's monitoring and emergency management efficiency for local vessels and typhoon shelters while greatly reducing labour and time costs.
The research results have been adopted by the Hong Kong Marine Department as a technical reference for local typhoon shelter planning, assisting the government in developing shelter facility plans and establishing benchmarks for the digital management of coastal vessels. The technology has broad applications in port state control inspections and port congestion management. Looking ahead, the team plans to develop techniques for data collection and processing based on videos and images, integrating deep learning algorithms to create an intelligent regulatory system for vessels and navigation areas. To date, the team has collected over 50,000 images of local vessels, laying a solid foundation for future research and analysis.
Utilising intelligent algorithms to calculate port congestion index
Accurately and timely obtaining maritime statistical indicators, such as port congestion index and shipping line connectivity index, remains a major challenge for the maritime industry. Traditional manual data collection methods are error-prone, and the macro information released by administrative agencies or port departments often suffers from delays and lacks comprehensiveness.
To address this, Prof. Yang Dong and his team have collaborated with researchers from Tsinghua University to develop advanced big data analysis algorithms for processing AIS data. They constructed a global, multi-level shipping and trade network database and created an online platform capable of calculating key indicators in real-time, such as port congestion index, port connectivity index, and port turnover rate. This research substantially broadens the application scenarios of maritime big data technologies, enabling the current shipping analysis from micro to macro levels, accurately capturing industry dynamics, and providing a reliable basis for maritime operational analysis and decision-making, thereby supporting the sustainable development of the industry.
Analysing trajectories to identify illegal fishing vessels
In the past, combating illegal fishing primarily relied on random sea patrol, which is costly and inefficient. To enhance the management efficiency of fishing vessels in Hong Kong waters, Prof. Yang Dong and his team developed a fishing vessel behaviour pattern recognition model by applying a novel trajectory feature engineering method combined with a semi-supervised machine learning framework. This model effectively identifies abnormal fishing behaviours with an accuracy rate of up to 90%.
This technology integrates maritime domain knowledge with AI algorithms to establish precise distinctions in vessel trajectory features under different navigation states. It greatly reduces the time and labour required for manual data collection and labelling process and can be flexibly adapted for vessel trajectory prediction and emissions monitoring. The team has collaborated with the Hong Kong Tourism Commission and the Marine Department to assess the risks associated with large cruise ships navigating Hong Kong's central channel using multi-source maritime data such as AIS, maritime radar, and CCTV footage. Recently, the team have utilised graph neural networks to accurately predict the future trajectories of multiple vessels in Hong Kong's busy waterways, further strengthening real-time navigation safety supervision.
Prof. Yang Dong said, "AI and big data technologies are bringing revolutionary changes to the maritime and shipping industry. These innovative assessment and monitoring tools combine domain knowledge in the maritime field with cutting-edge technology, significantly improving the speed, quality, and accuracy of data collection. They also address key operational challenges faced by the industry and make substantial contributions to academic research in related fields, promoting the intelligent development of maritime operations and further solidifying Hong Kong's position as an international maritime centre."
