Research on an Optimized Method for Collaborative Observation of Maritime Targets Based on Data Fusion

To address critical issues in traditional maritime target observation systems—such as frequent observation blind spots, insufficient data integrity, and delayed emergency responses caused by reliance on single platforms—this paper proposes an integrated optimization method that combines trajectory prediction, intelligent blind spot identification, multi-platform collaborative observation, and emergency response. This method first collects data via maritime observation platforms to construct target temporal trajectory maps. Long Short-Term Memory (LSTM) networks are employed to achieve high-precision prediction of target trajectories for the next 10–30 min. Based on trajectory predictions and 3D geographic modeling technology, a blind zone scanning system is established to accurately identify observation blind spots and generate spatiotemporal dynamic distribution maps. For short-term local and long-term wide-area blind spots, a layered collaborative compensation observation network integrating “UAVs - BeiDou satellites - shore-based platforms” is constructed. Scale-Invariant Feature Transform (SIFT) and weighted fusion algorithms achieve spatio-temporal alignment and pixel-level fusion of multi-source observation data. Finally, a decision tree model enables target anomaly behavior recognition and emergency response triggering. Experimental results demonstrate that this method elevates maritime target observation coverage to 96.5%, achieves a mean absolute error (MAE) ≤ 5m for trajectory prediction, and reduces emergency response time to within 30 s. Significantly outperforming traditional observation methods, it provides reliable technical support for maritime target monitoring and emergency management.