A Quasi-Intelligent Maritime Route Extraction from AIS Data.

Area of Science:

• Maritime technology
• Data science
• Machine learning

Background:

• Automatic Identification Systems (AIS) are increasingly used for maritime surveillance and route planning.
• Traditional manual route planning is subjective, experience-dependent, and time-consuming.
• There is a need for objective and efficient methods for extracting shipping routes.

Purpose of the Study:

• To extract shipping routes from historical AIS data using unsupervised machine learning algorithms.
• To construct a quantitative maritime traffic network reflecting ship characteristics.
• To evaluate the effectiveness of the proposed approach for route planning.

Main Methods:

• A three-step approach was developed: maneuvering point detection, waypoint discovery, and traffic network construction.
• Historical AIS data was utilized to build the maritime traffic network.
• The Symmetrized Segment-Path Distance (SSPD) metric was employed for comparison.

Main Results:

• The constructed maritime traffic network quantitatively reflects ship characteristics (length, type).
• The proposed approach successfully generated a maritime traffic network.
• Lower SSPD values indicated a close resemblance between the constructed network and actual ship transit routes.

Conclusions:

• The unsupervised machine learning approach is effective for extracting shipping routes from AIS data.
• The developed method provides a quantitative and data-driven approach to maritime route planning.
• This technique can enhance the efficiency and objectivity of maritime surveillance and planning.