Sea Surface Small Target Tracking Method with Integrated Dynamic Wave Prediction Information

Tracking small targets on the sea surface has always been a challenging issue in the field of maritime target monitoring. Traditional methods typically rely on specific dynamic equations to establish the motion state model of the target (such as CV, CA, and CT models or their combinations), which struggle to accurately represent the motion states of small targets under complex sea conditions. We propose a method that incorporates deep learning to predict the dynamic fluctuations of sea waves, integrating real-time wave state information into the target state model. This method, which fuses sea wave dynamic prediction information, aims to improve the tracking of small sea surface targets. Firstly, we simulate various sea conditions based on the Elfouhaily wave spectrum to establish a database of small target motion trajectories on the sea surface. Then, we employ an LSTM network with a dual-stage attention mechanism to predict wave heights, which are subsequently fed into the state transition model. We utilize the Extended Kalman Filter for tracking. Simulation results demonstrate that with the assistance of sea wave dynamic prediction information, the proposed method enhances target tracking accuracy by approximately 14% compared to traditional methods.