基于深度学习的声呐图像目标检测系统

Sonar Image Object Detection System Based on Deep Learning

摘要:

声呐图像目标检测是实现水下勘探、海底救援、敌对目标侦查等任务的重要环节，深度学习相关技术的突破为该领域的发展带来了新的机遇。基于深度学习的声呐图像目标检测算法性能优于传统方法，然而相关的系统性研究与应用仍然不足。鉴于此，利用深度学习模型数据驱动的优势设计了一种声呐图像目标检测系统，以满足实际应用对系统精度、速度、可移植性、可扩展性、部署环境的需求。该系统由数据集生成、算法模型训练与测试、模型部署应用三个子系统组成，应用于水下可疑目标探测任务，实验结果表明所实现的目标检测系统在测试数据上和实际应用中均具有良好的性能。

Abstract:

Sonar image object detection is an important part of underwater exploration, submarine rescue, hostile target reconnaissance and other tasks. The breakthrough of deep learning related technologies has brought new opportunities for the development of this field. The performance of sonar image object detection algorithm based on deep learning is better than traditional methods, but the relevant systematic research and application are still insufficient. Therefore, a sonar image object detection system is designed to meet the requirements of accuracy, speed, portability, extensibility, and deployment environment of the system in practical applications by taking advantage of the data driven advantages of the deep learning model. The system consists of three subsystems: data set generation, algorithm model training and testing, and model deployment application, which is applied to the underwater suspicious target detection task. The experimental results show that the object detection system has good performance in the test data and practical applications.