A Hierarchical Spatial-Temporal Embedding Method Based on Enhanced Trajectory Features for Ship Type Classification.

Area of Science:

• Maritime Navigation
• Data Science
• Machine Learning

Background:

• Ship type classification is crucial for maritime monitoring and analysis.
• Existing trajectory-based methods struggle with unevenly distributed geographical data.
• Accurate ship movement pattern inference is challenging with raw trajectory data.

Purpose of the Study:

• To propose a novel method for ship type classification using enhanced trajectory features.
• To address the limitations of existing methods in handling real-world, uneven ship trajectory data.
• To improve the accuracy and efficiency of ship classification in maritime domains.

Main Methods:

• Trajectory preprocessing and integration with port information to create moored records.
• Development of the Hierarchical Spatial-Temporal Embedding Method (Hi-STEM).
• Mapping moored records into a feature space for efficient classification plane identification.

Main Results:

• The proposed Hi-STEM method achieves high accuracy in ship type classification.
• The approach effectively handles unevenly distributed trajectory points by focusing on moored records.
• Experimental results demonstrate superior performance compared to existing methods on real-world datasets.

Conclusions:

• The Hi-STEM method offers a robust solution for ship type classification using moored records.
• Enhanced trajectory features and spatial-temporal embedding significantly improve classification accuracy.
• The publicly available code and datasets facilitate further research and application in maritime surveillance.