Vessel Trajectory Anomaly Detection Based on Multi-scale Convolutional Autoencoder

Traditional methods for detecting anomalies in vessel trajectories do not adequately account for the multidimensional characteristics of vessel behavior, and lack the ability to capture trajectory details. To address these issues, a method of vessel trajectory anomaly detection based on Multi-scale...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Ocean engineering Jg. 343; S. 123564
Hauptverfasser: Qi, Yuhao, Yang, Jiaxuan, Xu, Dongsheng, Shao, Ran, Duan, Yangyang, Wang, Yangjie
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier Ltd 15.01.2026
Schlagworte:
Anomaly detection
Multi-scale convolutional autoencoder
Multidimensional features
Vessel trajectory
Multi-scale convolutional autoencoder
Vessel trajectory
Multidimensional features
Anomaly detection
ISSN:0029-8018
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Traditional methods for detecting anomalies in vessel trajectories do not adequately account for the multidimensional characteristics of vessel behavior, and lack the ability to capture trajectory details. To address these issues, a method of vessel trajectory anomaly detection based on Multi-scale Convolutional Autoencoder (MCAE) is proposed, which fully leverages the multidimensional behavioral features of vessels to enhance the precision of anomaly detection. Firstly, a mechanism for generating multidimensional features information trajectory maps is presented. Secondly, an MCAE model capable of effectively capturing trajectory details is constructed. Building upon the autoencoder, local and global features of trajectory maps at various resolutions are extracted through multi-scale convolution and fused into intermediate decoder layers by skip connections, achieving an enhanced ability to capture trajectory details. Finally, an algorithm of vessel trajectory anomaly detection based on MCAE is proposed. The Structural Similarity Index Measure (SSIM) is applied to calculate anomaly scores for maps reconstructed by the model, and a threshold is set to achieve the anomaly detection of vessel trajectories. The Port of Tianjin is used as the study area, and experimental results demonstrate that the proposed method effectively detects anomaly vessel trajectories, closely matching actual conditions. Compared to other methods, it shows superior reconstruction accuracy and detection precision, offering robust support for port security management and maritime traffic safety. •A mechanism for generating multidimensional features trajectory maps is presented.•An MCAE model capable of effectively capturing trajectory details is constructed.•An algorithm of vessel trajectory anomaly detection based on MCAE is proposed.•The proposed method is verified through real case analysis and model comparison.
ISSN:0029-8018
DOI:10.1016/j.oceaneng.2025.123564