Ships Classification Using Deep Neural Network Based on Attention Mechanism

Abstract

As a serviceable tool of underwater targets classification for sonar operators, deep neural network behaves a good work on underwater targets intelligent classification. Since the line frequencies of radiated noise supplied distinct frequency bands for different ships, a deep neural network is proposed based on an attention mechanism to improve the classification accuracy in this paper. The results show that the equilibrium classification accuracy of ACNN-QJ4 is 9.15% higher than that of non-attention network. Finally, by comparing the output features of these two networks with and not with attention mechanism, the superiority on feature extracting of the attention network proposed by this paper has been shown.