Abstract:The identification of atypical vessel behavior is a crucial factor in ensuring the safety of maritime navigation and enhancing the effectiveness of waterway oversight. The current approach to detecting vessel abnormal behavior lacks consideration for the inter-dependency among different time series and various characteristics associated with abnormal vessel behavior. Additionally, the subjective setting of anomaly detection threshold adversely affects the real-time performance of anomaly detection, resulting in poor detection accuracy. In order to tackle these issues, this paper explores the monitoring and detection mechanism of vessel abnormal behavior based on Graph Attention Prediction and Reconstruction Network. The proposed mechanism employs a sliding window to generate fixed data input. It effectively captures the inter-dependencies among vessel behavior characteristics. Additionally, it dynamically adjusts the threshold for identifying anomalies based on variations in the marine environment, utilizing the Peak Over Threshold (POT) method. In order to achieve real-time detection of abnormal vessel behavior in multivariate, multi-dimensional time series data, it is necessary to complete the task. The results of the paper are derived from an analysis of the AIS dataset encompassing vessel movements between Dalian and Yantai. The temporal scope of the dataset spans from November 1, 2021, to December 31, 2021. The detection accuracy, detection precision and recall rate are 98.29%, 92.44%, 87.09% respectively. The results indicate that there is an enhancement in the accuracy of detecting abnormal behavior in vessels, leading to a significant reduction in the blind area of VTS (Vessel traffic services) system monitoring.

Comparative evaluation of VAE-based monitoring statistics for real-time anomaly detection in AIS data

Robust and Unsupervised KPI Anomaly Detection Based on Highly Sensitive Conditional Variational Auto-Encoders.

Systematic Development of a New Variational Autoencoder Model Based on Uncertain Data for Monitoring Nonlinear Processes

Situation-Aware Multivariate Time Series Anomaly Detection Through Active Learning and Contrast VAE-Based Models in Large Distributed Systems

An Ensemble-Based Approach to Anomaly Detection in Marine Engine Sensor Streams for Efficient Condition Monitoring and Analysis

Variational Autoencoder for Anomaly Detection: A Comparative Study

Machine Learning-Assisted Anomaly Detection in Maritime Navigation Using AIS Data

Self-Supervised Variational Graph Autoencoder for System-Level Anomaly Detection

A Novel Model for Ship Trajectory Anomaly Detection Based on Gaussian Mixture Variational Autoencoder

Unsupervised maritime anomaly detection for intelligent situational awareness using AIS data

A novel process monitoring approach based on variational recurrent autoencoder.

Enhancing Anomaly Detection in Maritime Operational IoT Time Series Data with Synthetic Outliers

Unsupervised Anomaly Detection for Intricate KPIs via Adversarial Training of VAE

Multivariate time series anomaly detection with variational autoencoder and spatial–temporal graph network

MA-VAE: Multi-head Attention-based Variational Autoencoder Approach for Anomaly Detection in Multivariate Time-series Applied to Automotive Endurance Powertrain Testing

The model of vessel trajectory abnormal behavior detection based on graph attention prediction and reconstruction network

TeVAE: A Variational Autoencoder Approach for Discrete Online Anomaly Detection in Variable-state Multivariate Time-series Data

Anomaly Detection in Maritime AIS Tracks: A Review of Recent Approaches

Distributed temporal–spatial neighbourhood enhanced variational autoencoder for multiunit industrial plant‐wide process monitoring

Toward a Better Monitoring Statistic for Profile Monitoring via Variational Autoencoders

Real-Time Predictive Condition Monitoring Using Multivariate Data