第15卷第3期 智能系统学报 Vol.15 No.3 2020年5月 CAAI Transactions on Intelligent Systems May 2020 D0L:10.11992tis.202004026 海底声呐图像智能底质分类技术研究综述 赵玉新，赵廷 (哈尔滨工程大学自动化学院，黑龙江哈尔滨150001) 摘要：海底声呐图像底质分类技术是指利用多波束、侧扫声呐等设备获取海底图像进行浅表层地质属性信息 的反演和预测。综合运用水声学、图像处理以及机器学习的相关理论，声学海底底质分类技术能够快速、淮确 地识别海底底质类型。通过回顾国内外发展历程，对利用声学图像进行海底底质分类的关键技术进行了总结， 从声学海底底质分类的关系模型、海底声呐图像的特征表达和分类模型构建三个方面介绍了领域内的研究进 展和主要方法，重点分析了不同模型和方法的原理、技术特点和适用场合，并结合卷积神经网络对深度学习方 法在海底底质分类中的应用进行了讨论。最后，对海底声呐图像底质分类技术的研究方向和发展趋势进行了 归纳和展望。 关键词：声学探测：声呐图像：底质类型：特征提取：图像分类：监督学习：无监督学习：深度学习：卷积神经网 络：海底底质分类 中图分类号：TP753文献标志码：A文章编号：1673-4785(2020)03-0587-14 中文引用格式：赵玉新，赵廷.海底声呐图像智能底质分类技术研究综述.智能系统学报，2020,15(3)：587-600. 英文引用格式：ZHAO Yuxin,,ZHAO Ting.Survey of the intelligent seabed sediment classification technology based on sonar im- ages[J].CAAI transactions on intelligent systems,2020,15(3):587-600. Survey of the intelligent seabed sediment classification technology based on sonar images ZHAO Yuxin,ZHAO Ting (College of Automation,Harbin Engineering University,Harbin 150001,China) Abstract:Image-based acoustic seabed sediment classification refers to the technology of inversion and prediction of the marine geological attributes of the shallow strata using seabed sonar image obtained using a multi-beam,side-scan sonar.As the multidisciplinary branch of oceanology,this technology is able to quickly identify a sediment type based on the relevant knowledge of underwater acoustics,image processing,and machine learning.Based on the review on the history and development of the technology at home and abroad,this article summarizes the key techniques in the frame- work of seabed sediment classification using sonar image and makes an introduction to the progress in research and main algorithms used in the domain,including the geoacoustic relationship model,the feature expression of the seabed sonar image,and the building of classification model.The emphasis is put on the analysis of the principles,technical features,and applications for various models and algorithms.Deep learning is also discussed for exploring proper ap- plication in the acoustic seabed classification with the case of convolutional neural network.The deep learning al- gorithms are applied on the sonar images and analyzed.Finally,acoustic image-based seabed sediment classification is completed and forecast is done. Keywords:acoustic detection;sonar image;sediment type;feature extraction;image classification;supervized learning; unsupervized learning:deep learning;convolutional neural network;seabed sediment classification 收稿日期：2020-04-24. 随着人类海上活动空间的进一步拓展，海洋 基金项目：国家重点基础研究发展计划(613317)：国家自然科 学基金面上项目(41676088). 矿产资源勘探、地球生命系统演化等一系列重要 通信作者：赵廷.E-mail:zhaoting@hrbeu.edu.cn. 的科学问题都对海底科学的发展提出了新的要DOI: 10.11992/tis.202004026 海底声呐图像智能底质分类技术研究综述 赵玉新，赵廷 （哈尔滨工程大学 自动化学院，黑龙江 哈尔滨 150001） 摘 要：海底声呐图像底质分类技术是指利用多波束、侧扫声呐等设备获取海底图像进行浅表层地质属性信息 的反演和预测。综合运用水声学、图像处理以及机器学习的相关理论，声学海底底质分类技术能够快速、准确 地识别海底底质类型。通过回顾国内外发展历程，对利用声学图像进行海底底质分类的关键技术进行了总结， 从声学海底底质分类的关系模型、海底声呐图像的特征表达和分类模型构建三个方面介绍了领域内的研究进 展和主要方法，重点分析了不同模型和方法的原理、技术特点和适用场合，并结合卷积神经网络对深度学习方 法在海底底质分类中的应用进行了讨论。最后，对海底声呐图像底质分类技术的研究方向和发展趋势进行了 归纳和展望。 关键词：声学探测；声呐图像；底质类型；特征提取；图像分类；监督学习；无监督学习；深度学习；卷积神经网 络；海底底质分类 中图分类号：TP753 文献标志码：A 文章编号：1673−4785(2020)03−0587−14 中文引用格式：赵玉新, 赵廷. 海底声呐图像智能底质分类技术研究综述 [J]. 智能系统学报, 2020, 15(3): 587–600. 英文引用格式：ZHAO Yuxin, ZHAO Ting. Survey of the intelligent seabed sediment classification technology based on sonar im￾ages[J]. CAAI transactions on intelligent systems, 2020, 15(3): 587–600. Survey of the intelligent seabed sediment classification technology based on sonar images ZHAO Yuxin，ZHAO Ting (College of Automation, Harbin Engineering University, Harbin 150001, China) Abstract: Image-based acoustic seabed sediment classification refers to the technology of inversion and prediction of the marine geological attributes of the shallow strata using seabed sonar image obtained using a multi-beam, side-scan sonar. As the multidisciplinary branch of oceanology, this technology is able to quickly identify a sediment type based on the relevant knowledge of underwater acoustics, image processing, and machine learning. Based on the review on the history and development of the technology at home and abroad, this article summarizes the key techniques in the frame￾work of seabed sediment classification using sonar image and makes an introduction to the progress in research and main algorithms used in the domain, including the geoacoustic relationship model, the feature expression of the seabed sonar image, and the building of classification model. The emphasis is put on the analysis of the principles, technical features, and applications for various models and algorithms. Deep learning is also discussed for exploring proper ap￾plication in the acoustic seabed classification with the case of convolutional neural network. The deep learning al￾gorithms are applied on the sonar images and analyzed . Finally, acoustic image-based seabed sediment classification is completed and forecast is done. Keywords: acoustic detection; sonar image; sediment type; feature extraction; image classification; supervized learning; unsupervized learning; deep learning; convolutional neural network; seabed sediment classification 随着人类海上活动空间的进一步拓展，海洋 矿产资源勘探、地球生命系统演化等一系列重要 的科学问题都对海底科学的发展提出了新的要 收稿日期：2020−04−24. 基金项目：国家重点基础研究发展计划 (613317)；国家自然科 学基金面上项目 (41676088). 通信作者：赵廷. E-mail：zhaoting@hrbeu.edu.cn. 第 15 卷第 3 期 智 能 系 统 学 报 Vol.15 No.3 2020 年 5 月 CAAI Transactions on Intelligent Systems May 2020