The sea surface escort formation faces various threats in reality. For example, suicide boats may carry explosives or other dangerous items, aiming to cause maximum damage by colliding or detonating escort targets. Since suicide boats have a certain degree of concealment, it is necessary to establish a threat assessment algorithm to timely identify and respond to such fast and concealed threats. This paper establishes a threat assessment model that considers the instantaneous and historical states of the target. The instantaneous state of the target takes into account six evaluation indicators, including target category, target distance, target heading, target speed, collision risk, and ship automatic identification system (AIS) recognition status; in terms of historical state information mining, a target typical intention recognition method based on graph neural network is proposed to achieve end-to-end target typical intention recognition. Furthermore, this paper introduces a multi-attribute decision analysis method to weight the evaluation indicators, improves the relative closeness calculation method between different evaluation schemes and positive and negative ideal schemes, and determines the target threat ranking based on relative closeness. Based on Unity3D, a set of unmanned boat confrontation simulation system is designed and developed, and typical intention recognition data sets and threat assessment scenario simulation data are generated through real-life confrontation. Comparative analysis shows that the threat assessment model in this paper can accurately and timely detect raid target threats and give scientific and reasonable target threat ranking results.
You may also start an advanced similarity search for this article.