Advances in Underwater Target Passive Recognition Using Deep Learning

In recent years, with the great breakthroughs in the theory of deep learning, it shows obvious advantages over traditional machine learning methods. By the remarkable feature learning ability, deep learning has been successfully applied to speech and image recognition, and it quickly attracted the attention of researchers in other fields. This paper summarizes the current research progress of deep learning based underwater target passive recognition methods, including the popular architectures of deep neural networks, the changes in feature extraction methods and modeling method under the condition of insufficient data. In order to meet the challenges in the future, this paper also prospects some possible research directions, which can be used for reference by relevant researchers.