Ship behavior prediction and anomaly detection using LSTM-DCross model based on AIS and remote sensing data
Accurate and reliable prediction of ship movements is critical for maritime safety and traffic management, yet it remains challenging due to the nonlinearity, stochasticity, and trendiness of ship motions. This paper proposes a deep cross LSTM network (LSTM-DCross) based on AIS time series data to o...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Taylor & Francis Group
2025-08-01
|
| Series: | International Journal of Digital Earth |
| Subjects: | |
| Online Access: | https://www.tandfonline.com/doi/10.1080/17538947.2025.2515251 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Accurate and reliable prediction of ship movements is critical for maritime safety and traffic management, yet it remains challenging due to the nonlinearity, stochasticity, and trendiness of ship motions. This paper proposes a deep cross LSTM network (LSTM-DCross) based on AIS time series data to overcome limitations in traditional approaches. The model integrates an encoder-decoder framework with an attention mechanism to predict ship behaviours, including position (longitude, latitude), course (COG), and speed (SOG). The study addresses two critical scenarios: GPS failure and complete AIS failure, utilizing two prediction methods – joint prediction and historical data-based prediction – to forecast the target ship behaviour. Experimental results show that LSTM-DCross outperforms BP, LSTM, LSTM-ATT, and TCN-GRU networks, achieving an MSE of 0.0754 and MAE of 0.1891. Notably, the historical data-based prediction method with LSTM-DCross not only demonstrates high accuracy but also offers greater applicability, successfully predicting future ship behaviours using only historical data, even during complete AIS failure. Additionally, by comparing AIS data with LSTM-DCross predictions, deviation values are calculated to detect abnormal ship behaviours. When deviations exceed a predefined threshold, the system records data and issues warnings, providing a potential tool for enhanced ship monitoring and risk mitigation. |
|---|---|
| ISSN: | 1753-8947 1753-8955 |