Underwater Target Maneuver Detection and Motion Mode Recognition Based on Bi-LSTM
Yan Zhu, Weining Mao · 2024
Traditional bearing-only maneuver detection methods for underwater targets often struggle with distinguishing the specific maneuver mode, are susceptible to observation noise, and show limited accuracy for subtle maneuvers. To address these shortcomings, this paper proposes a method for underwater target maneuver detection and motion mode recognition based on a bidirectional long short-term memory (Bi-LSTM) classification neural network. By establishing the nonlinear mapping relationship between bearing and range observations and the motion mode of the target, this method can recognize uniform linear motion and three kinds of maneuver: heading change, velocity change, and a combination of both heading and velocity changes. Then maneuver detection is further realized by analyzing the classification results of the target motion mode. As demonstrated by simulation results, this method maintains high accuracy in maneuver detection and motion mode recognition, even under significant observational noise or when faced with a subtle maneuver.