Answers are grounded in this paper and cite their sources.
Monitoring vessels in complex waterways using UAV-based maritime surveillance faces severe accuracy degradation due to dynamic spatio-temporal misalignment between AIS and video data, fluctuating data quality, and trajectory discontinuity during AIS outages. To address these challenges, an AI-driven, integrated fusion framework is proposed for simultaneously reconstructing trajectories and recognizing vessel identities. The method leverages real-time 6-DoF pose measurements from the onboard GNSS/IMU to construct a dynamic spatio-temporal reference, and incorporates an adaptive multi-feature similarity metric with a confidence-weighted fusion strategy to handle asynchronous multi-modal matching and data quality fluctuations. A video-assisted Adjusted Kinematic Interpolation (AKI) scheme compensates for missing AIS data, forming a closed loop of alignment–association–reconstruction. In a 6-day UAV-based maritime surveillance experiment on the Yangtze River in China, the proposed method reduces trajectory RMSE by 56.7% compared to pure-AIS interpolation and by 27.6% compared to shore-based video assistance. The MOFA fusion accuracy is 7.3% higher than MSF-VIT and 8.2% higher than Sigmoid weighting. The F1-score for vessel identity recognition is 7.8% higher than MSF-VIT. These results demonstrate that the proposed AI-enhanced method effectively resolves challenges of trajectory continuity and identity maintenance in dynamic UAV cruise scenarios, providing a robust solution for intelligent UAV-based maritime surveillance.
No takes yet. Share an insight, caveat, or question.
Yang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: