摘要
海浪有效波高是近岸海洋观测的重要要素,近岸摄像头拍摄的图像可直观地反映波高大小,但目前基于图像的有效波高反演算法研究多处于室内实验阶段且方法缺乏普适性。本文基于深度学习技术,以山东青岛小麦岛近岸海浪为例,基于海浪图像和浮标实测数据,开展近岸海浪图像反演有效波高方法研究,给出一种利用图像反演海浪有效波高的方法,该方法利用卷积网络提取海浪图像的特征,利用全连接网络提取风速等气象特征,将特征融合后作为全连接层的输入,最后输出反演的有效波高。通过对比多种模型的反演结果和浮标观测数据,发现多参数DenseNET121模型有效波高反演能力优于其他神经网络模型,其平均绝对误差为8.97 cm。本文研究为近岸海浪观测提供了一种新的技术思路。
The significant wave height is an important factor of offshore ocean observation,and the images taken by the offshore camera can directly reflect the wave height,but the current researches on the significant wave height inversion algorithm based on image are mainly conducted for laboratory experiment and universal method is lacked.Based on the deep learning technology,taking the inshore wave of the Xiaomai Island in Qingdao,Shandong Province as an example,based on the wave image and buoy data,the method of inshore wave image inversion effective wave height is studied,and a method of image inversion effective wave height is given.In this method,multiple convolutional networks are used to extract the features of wave images,and full connection networks are used to extract meteorological features such as wind speed.The features are fused as the input of the full connection layer,and finally the effective wave height of the inversion is output.By comparing the inversion results of various models with the buoy observation data,it is found that the inversion ability of the multi-parameter Dense Net121 model is superior to other neural network models,and the average absolute error is 8.97 cm.The research in this paper provides a new technical idea for offshore wave observation.
作者
黄文华
胡伟
崔学荣
曾强胜
商杰
王宁
李锐
Huang Wenhua;Hu Wei;Cui Xuerong;Zeng Qiangsheng;Shang Jie;Wang Ning;Li Rui(North China Sea Marine Forecasting Center of State Oceanic Administration,Qingdao 266061,China;College of Oceanography and Space Informations,China University of Petroleum,Qingdao 266555,China)
出处
《中国海洋大学学报(自然科学版)》
CAS
CSCD
北大核心
2024年第6期35-44,共10页
Periodical of Ocean University of China
基金
国家重点研究发展计划项目“海洋动力灾害观测预警系统集成与应用示范”(2018YFC1407002)资助。