A Maritime Target Classification Method for ARPA Radar Fusing Global Attention Mechanism with Spatiotemporal Features

To address the limitation of traditional maritime target classification methods in extracting key spatiotemporal features from automatic radar plotting aid (ARPA) radar trajectory data, a maritime target classification method for ARPA radar fusing global attention mechanism and spatiotemporal features is proposed in this paper. First, aiming at the insufficient extraction of trajectory spatial features caused by the single-channel input structure, a trajectory spatial feature extraction method based on multi-channel input one-dimensional convolutional neural network is proposed to realize the effective extraction of trajectory spatial features. Then, to solve the difficulty of capturing long-term dependent information of trajectory data in the long short-term memory (LSTM) network, a trajectory temporal feature extraction method based on bidirectional LSTM is presented, which can effectively extract trajectory features over long time intervals in both past and future periods. Finally, in view of the problem that key features are weakened due to equal weight assignment when fusing spatial and temporal features, a global attention-based spatiotemporal feature fusion strategy is proposed to achieve dynamic fusion according to feature importance. Experiments are conducted using real ARPA data to verify the effectiveness of the proposed method. The experimental results show that the classification accuracy of the proposed method reaches 97.45% on the real ARPA dataset. Compared with three other commonly used classification methods, the accuracy of the proposed method is improved by 9.77 %, 5.75 % and 2.20 %, respectively. The research indicates that the proposed method can effectively solve the problem of maritime target classification and achieve precise classification of maritime targets using ARPA radar.