Ship navigational status classification based on the geometrical and spatiotemporal features of the AIS-generated trajectory
Wijaya, W.M.; Nakamura, Y. (2024). Ship navigational status classification based on the geometrical and spatiotemporal features of the AIS-generated trajectory, in: 2024 the 9th International Conference on Big Data Analytics (ICBDA), March 16-18, 2024 Tokyo, Japan. pp. 103-112. https://dx.doi.org/10.1109/icbda61153.2024.10607233
|
| Beschikbaar in | Auteurs |
|
Documenttype: Congresbijdrage
|
| Auteurs | | Top |
- Wijaya, W.M.
- Nakamura, Y.
|
|
|
| Abstract |
Studies on the Automatic Identification System (AIS) data mining have proven the feasibility of discovering knowledge from the AIS tracks to understand vessels' behavior. One of its purposes is to help maritime surveillance efforts for safety and security matters. Research on vessel behavior often relies on the spatiotemporal features of the AIS tracks, while human operators recognize vessel movement by comprehending their trajectory patterns. Additionally, many works do not comprehensively discuss a thorough AIS data preprocessing stage before implementing their novelty method to analyze raw AIS data. Thus, this paper proposes a trajectory segmentation technique, a data preprocessing procedure, for AIS data to provide an analytics-ready dataset that facilitates a practical implementation of AIS-based vessel behavior analytics. The approach utilizes the AIS-generated vessel trajectories' spatiotemporal and geometrical features to split a ship trajectory into segments according to its navigational status, such as stopping and underway. The spatiotemporal features are speed, course, heading, and position timestamp, while the geometrical features include the diameter and bounding box of the trajectory. Every trajectory is divided into stopping and underway segments. Stopping segment extraction aims to facilitate stopping event detection and classification, which helps understand vessel activities in port areas. Mean-while, the underway segment is dedicated to streamlining vessel movement analytics for loitering detection and other anomalies in maritime traffic. The experiment result demonstrates the efficacy of the method. It outperformed the existing stopping segments classification approach with 0.98 accuracy and improved the currently published loitering detection to 0.97 accuracy and 0.91 F-score. |
|
