基于计算兴波云图的舰船航速智能识别方法  

Intelligent identification method of vessel speed based on computational wave pattern

作　　者：黄文育[1] 孟斌[1] HUANG Wen-yu;MENG Bin(China Ship Scientific Research Center,Wuxi 214082,China)

机构地区：[1]中国船舶科学研究中心,江苏无锡214082

出　　处：《舰船科学技术》2023年第10期147-151,共5页Ship Science and Technology

摘　　要：数据挖掘是船舶领域中的研究热点,针对船舶领域流场分析问题,基于波形图的航速判断误差较大,提出一种人工智能解决方法。由于可利用的数据集较少且图像噪声较多,采用数据增广扩充数据。通过图像处理降噪与生成对抗网络增加仿造样本,结合深度神经网络迁移学习的方式进行模型融合并分类,分别通过原始样本实验测试、数据扩增实验测试和迁移学习实验测试并比较,得出该方法预测结果接近实际航速,在样本丰富的情况下可以实际应用。Data mining is a research hotspot in the field of ships.Aiming at the problem of flow field analysis in the field of ships,in which the speed judgment error based on the waveform diagram is large,an artificial intelligence solution is proposed.Since there are fewer available data sets and more image noise,it is necessary to use data augmentation to expand the data,through image processing to reduce noise and generate countermeasure networks to increase imitation samples,combined with deep neural network migration learning methods for model fusion and classification,respectively Through the original sample experiment test,the data amplification experiment test and the transfer learning experiment test and comparison,it is concluded that the prediction result of this method is close to the actual speed,which can be used in the case of abundant samples.

关 键 词：数据挖掘 波形图 生成对抗网络 迁移学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]