Related Experiment Video
Updated: Jul 14, 2025
06:37
Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
Published on: December 15, 2023
3.8K
Loitering behavior detection by spatiotemporal characteristics quantification based on the dynamic features of
Wayan Mahardhika Wijaya1, Yasuhiro Nakamura2
1Graduate School of Science and Engineering, National Defense Academy of Japan, Yokosuka, Kanagawa, Japan.
Peerj. Computer Science
|October 9, 2023
Summary
This study introduces a new, region-independent method to automatically detect ship loitering using Automatic Identification System (AIS) data. The system effectively identifies potentially anomalous vessel behavior, improving maritime surveillance efficiency.
Area of Science:
- Maritime surveillance
- Data science
- Naval technology
Background:
- Automatic Identification System (AIS) provides global ship tracking data for maritime surveillance.
- Detecting anomalies in ship behavior is vital for identifying emergencies and illegal activities.
- Manual examination of ship behavior is labor-intensive and inefficient, especially for loitering, a common but under-explored anomaly.
Purpose of the Study:
- To develop a region-independent computational method for automatically detecting loitering behavior in maritime traffic.
- To create a system that does not require training on normal instances and can rank suspicious vessels.
- To enhance the efficiency and accuracy of maritime anomaly detection, particularly for cargo and tanker vessels.
Main Methods:
- Defined spatiotemporal characteristics of loitering, including frequent course changes, speed, spatial range, and discrepancies between course over ground and heading.
- Quantified these characteristics using AIS dynamic information and formulated parameters to assess loitering trajectories.
- Employed the Isolation Forest algorithm for thresholding and ranking, coupled with geographic visualization for intuitive evaluation.
Main Results:
- The proposed method achieved 97% accuracy and a 92% F-score in detecting loitering on a large real-world dataset.
- Successfully produced a ranked list of loitering vessels and intuitive geographic visualizations.
- Demonstrated superior performance compared to existing approaches in identifying anomalous maritime behavior.
Conclusions:
- The developed region-independent method effectively automates the detection of loitering, a significant maritime anomaly.
- The system provides practical tools for human operators to support further investigation of suspicious vessels.
- This approach enhances the capability of maritime authorities to monitor and manage vessel traffic more efficiently.
