| 495 | 6 | 251 |
| 下载次数 | 被引频次 | 阅读次数 |
针对水体运动导致电子传感装置测量结果准确性下降的问题和阈值分割法无法在光照场景下测量波面的问题，文章提出了一种基于U-net卷积神经网络的波浪测量方法。实验过程首先由高清摄像机录制水槽中的波浪运动过程，将视频处理成时间连续的波面图像，其次通过U-net卷积神经网络对波面图像进行图像分割并提取水位线数据信息，最后求出波高和周期。以像素识别结果为基准，将本研究方法的测量结果与波高传感器的测量结果进行误差对比，结果表明U-net卷积神经网络的相对误差最大为2.25%,而传感器误差最大为4.15%,且实验组中U-net卷积神经网络测得平均波高的相对误差均在2.5%以内，平均周期的误差都低于1%。因此，基于U-net卷积神经网络的测量方法可用于实验室的波浪测量。
Abstract：Aiming at the problem that the water movement leads to the decrease of the accuracy of the measurement results of the electronic sensing device and the problem that the threshold segmentation method cannot measure the wave surface in the illumination scene. This paper proposes a wave measurement method based on U-net convolution neural network, which overcomes the impact of water movement on the measurement accuracy of electronic sensing devices and the deficiencies that the illumination from water surface has a great impact on the threshold segmentation. Firstly, the wave motion images are recorded by the high-definition camera in the water tank. Secondly, the U-net convolution neural network is used to segment the wave images, and then the water level data information is extracted. Finally, the wave height and period are obtained. Based on pixel recognition results, the maximum relative error of U-net method is 2.25%, which is smaller than that of the wave gauge with 4.15% maximum relative error. The results show that the relative errors of the average wave height are less than 2.5%, and the errors of the average period are within 1%. Therefore, the measurement method based on U-net convolution neural network can be used for laboratory wave measurement.
[1] 陈进，陈广来，陈汉宝，等.电容式波高传感器采样稳定性实验研究[J].天津理工大学学报，2016,32(1):10-13.Chen J,Chen G L,Chen H B,et al.Capacitive high wave sensors sampling stability experiment[J].Journal of Tianjin University of Technology,2016,32(1):10-13.
[2] 王晨阳，王收军，陈松贵，等.基于加速度计和陀螺仪的波浪测量方法[J].科学技术与工程，2019,19(6):44-49.Wang C Y,Wang S J,Chen S G,et al.Wave measurement method based on accelerometer and gyroscope[J].Science Technology and Engineering,2019,19(6):44-49.
[3] 赵刚，唐得刚.几种常用的液位在线检测方法的比较[J].中国仪器仪表，2005(5):36-38,40.Zhao G,Tang D G.The comparison between common methods of on-line measurement of liquid level[J].China Instrumentation,2005(5):36-38,40.
[4] 马腾飞，王收军，陈汉宝，等.基于光学的波浪测量方法在波态研究中的应用[J].水道港口，2017,38(3):308-312.Ma T F,Wang S J,Chen H B,et al.Application of wave measurement based on optical method in wave shape study[J].Journal of Waterway and Harbor,2017,38(3):308-312.
[5] 常亮，解建仓，等.应用优化神经网络算法预报地下水位[J].水利水运工程学报，2005(4):66-70.Chang L,Xie J C,et al.Prediction of groundwater level by optimized neural network algorithm[J].Hydro-Science and Engineering,2005(4):66-70.
[6] Kim J,Park C S,Min K W,et al.Fast vision-based wave height measurement for dynamic characterization of tuned liquid column dampers[J].Measurement,2016,89:189-196.
[7] 黄鹏，郑淇，梁超，等.图像分割方法综述[J].武汉大学学报(理学版),2020,66(6):519-531.Huang P,Zheng Q,Liang C,et al.Overview of image segmentation methods[J].Journal of Wuhan University(Natural Science Edition),2020,66(6):519-531.
[8] 侯红英，高甜，李桃，等.图像分割方法综述[J].电脑知识与技术，2019,15(5):176-177.Hou H Y,Gao T,Li T,et al.Overview of image segmentation methods[J].Computer Knowledge and Technology,2019,15(5):176-177.
[9] 肖飞，綦星光.图像分割方法综述[J].可编程控制器与工厂自动化，2009(11):77-79.Xiao F,Qin X G.A survey of image segmentation[J].Programmale Controller and factory automation,2009(11):77-79.
[10] 于宁波，刘嘉男，高丽，等.基于深度学习的膝关节MR图像自动分割方法[J].仪器仪表学报，2020,41(6):140-149.Yu N B,Liu J N,Gao L,et al.Auto-segmentation method based on deep learning for the knee joint in MR images[J].Chinese Journal of Scientific Instrument,2020,41(6):140-149.
[11] 陆展鸿，单鲁斌，苏立循，等.基于U-Net的掌纹图像增强与ROI提取[J].北京航空航天大学学报，2020,46(9):1807-1816.Lu Z H,Shan L B,Su L X,et al.Palmprint enhancement and ROI extraction based on U-Net[J].Journal of Beijing University of Aeronautics and Astronautics,2020,46(9):1807-1816.
[12] Ronneberger O,Fischer P,Brox T.U-Net:Convolutional networks for biomedical image segmentation[C]// Bildverarbeitung für die Medizin.Heidelberg:Algorithmen Systeme Anwendungen,2015.
[13] 王谭，王磊磊，张卫国，等.基于张正友标定法的红外靶标系统[J].光学精密工程，2019,27(8):1828-1835.Wang T,Wang L L,Zhang W G,et al.Design of infrared target system with Zhang Zhengyou calibration method[J].Optics and Precision Engineering,2019,27(8):1828-1835.
[14] 谭俊杰，钟妤，黄泽斌，等.基于U-net的直肠癌智能分割[J].计算机时代，2020(8):18-20,26.Tan J J,Zhong Y,Huang Z B,et al.Intelligent segmentation of rectal cancer based on U-net[J].Computer Era,2020(8):18-20,26.
[15] 马明明，姜原，刘义，等.U-Net深度学习模型对DCE-MRI上乳腺肿块自动分割和定位的准确性分析[J].放射学实践，2020,35(8):1030-1036.Ma M M,Jiang Y,Liu Y,et al.Automatic breast cancer segmentation and location on DCE-MRI using U-Net convolution neural network:A preliminary study[J].Radiologic Practice,2020,35(8):1030-1036.
[16] 郑艳梅，黄光耀，芦碧波，等.结合深度学习和引导滤波的苹果叶片图像分割[J].中国农业大学，2020,25(9):112-120.Zheng Y M,Huang G Y,Lu B B,et al.An apple leaf image segmentation algorithm using deep learning and guided filtering[J].Journal of China Agricultural University,2020,25(9):112-120.
基本信息:
DOI：10.16441/j.cnki.hdxb.20210195
中图分类号:TP183;TP391.41;P731.22;P753
引用信息:
[1]任志伟,陈松贵,王收军,等.基于U-net卷积神经网络图像分割的波浪测量方法[J],2022,52(09):125-132.DOI:10.16441/j.cnki.hdxb.20210195.
基金信息:
国家自然科学基金项目(52001149,52039005,51861165102);; 中国科协青年人才托举工程(2018QNRC001);; 中央级公益性科研院所基本科研业务费(TKS20200402,TKS20200204,TKS20200203);; 天津市科技计划项目(17PTYPHZ00080)资助~~
2021-04-22
2021
2021-05-20
2021-05-28
2021
1