A Position Probability Prediction Method of Marine Moving Targets for Optimization Search by Satellite

Marine moving ship target search has great application demand and development prospects in marine surveillance, marine search and rescue, customs anti-smuggling, and other aspects. Satellites are a vital means of searching for marine moving targets. Ground coverage for imaging satellites is relatively limited; the vast marine area makes searching for marine moving ships challenging. Target position prediction, used as the guidance information for satellite detection to reduce the uncertainty of the target position and improve search efficiency, is an essential approach to achieving target optimization search. This paper proposes a target position prediction method based on joint learning of multi-source factors for optimal satellite search of marine moving targets. Firstly, it analyzes the characteristics of marine moving targets and the factors affecting motion. Then, it puts forward the motion prediction model based on a deep neural network, integrating multifactor, like the state of the moving target, environmental attributes such as nautical charts, meteorology and hydrology, and kinematic constraints, to predict the target positional distribution probability. Finally, the target search strategy based on the maximum sum of the distribution probabilities of the satellite coverage area is adopted to improve target detection probability. The experiment results show that the prediction method can realize the target positional distribution probability prediction, reduce the uncertainty search when making the satellite search plan, and improve efficiency.