Deep Learning-Driven Sonar Image Fish Species Recognition and Behavioral Feature Analysis
Yao Jin（Corresponding author）  Feng Jing

School of mechanical and resource engineering, Wuzhou university, Guangxi Wuzhou 543000；

Abstract：In fields such as aquatic ecology research, fisheries management, and underwater monitoring, sonar imaging technology has become a key method for acquiring fish information due to its insensitivity to water quality and light interference. However, traditional analysis methods rely on manual expertise, suffering from low efficiency, poor accuracy, and strong subjectivity, making it difficult to meet large-scale real-time demands. The powerful feature extraction and pattern recognition capabilities of deep learning offer a new pathway to address these challenges.

This paper focuses on the application of deep learning in sonar image fish recognition and behavior analysis. It begins by explaining the imaging principles and characteristics of sonar images, analyzing the limitations of traditional methods. It then reviews models suitable for sonar image processing, such as Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and Transformers, discussing their application logic in object detection, classification, and behavior extraction. Subsequently, targeting key issues like target blurring, noise interference, and target occlusion, it proposes model optimization and data augmentation strategies. Finally, considering practical scenarios,Outlook for the application prospects of this technology in fisheries assessment, endangered species protection, underwater ecological monitoring, and other areas, providing references for related practices and research.

Keywords：Behavioral Feature Analysis; Object Detection; Feature Extraction

References

[1] Ship and Ocean Structure Design and Manufacturing. Deep Learning-Based Method for Small Underwater Target Recognition in Sonar Images [D]. 2023.

[2] Hu Gang. Research on Underwater Target Recognition and Motion Behavior Analysis Technology Based on Deep Learning [D]. Harbin Engineering University, 2021.

[3] Pattern Recognition and Intelligent Systems. Research on Underwater Target Recognition and Motion Behavior Analysis Technology Based on Deep Learning [D]. 2021.

[4] Zhou Qiuru. Research on Sonar Image Target Detection and Data Augmentation Methods Based on Deep Learning [D]. Harbin Engineering University, 2023.

[5] Xu Huanyu. Research on Classification of Forward-Looking Sonar Images Based on Deep Learning [D]. Harbin Engineering University, 2020.