Cross-Modal AI Framework Improves Vessel Trajectory Prediction
Researchers propose CmIVTP, a cross-modal interaction-based framework for vessel trajectory prediction in maritime intelligent transportation systems. The framework integrates AIS and CCTV data to overcome limitations of single-source data, which is often sparse or unavailable for small vessels. It introduces a target-aware scene encoder to extract semantic features and a cross-modal interaction transformer to model vessel-environment interactions, enhancing prediction accuracy.
Key facts
- CmIVTP stands for Cross-modal Interaction-based Vessel Trajectory Prediction
- The framework is designed for maritime intelligent transportation systems
- It addresses challenges of sparse AIS data and limited CCTV data
- A target-aware scene encoder extracts scene semantic features
- A cross-modal interaction transformer integrates AIS motion features
- The research is published on arXiv with identifier 2605.26524
- The submission type is cross
- The framework models interactions between vessel dynamics and environmental constraints
Entities
Institutions
- arXiv