Maritime target intent recognition based on Transformer

Maritime target intent recognition refers to the process of inferring the intent and objectives of maritime targets through the analysis of their behavior, movement trajectories, and communications. This recognition is crucial for maritime security and defense. Currently, many methods for inferring tactical intent employ recurrent neural networks (RNNs) and their variants. However, due to the relatively slow movement of maritime targets, similar patterns may be observed over short time sequences, necessitating a longer duration to accurately reflect the target’s intent. RNNs tend to suffer from memory degradation on long-sequence data, leading to the loss of early data and a decline in the accuracy of intent recognition for maritime targets. To address this issue, this paper proposes a Transformer-based intent recognition model, which demonstrates improved capability for capturing long-term dependencies and effectively mitigates long-term memory degradation. The efficacy of the proposed model is validated through comparative experiments on a maritime dataset obtained from simulations.