[1]何盛琪,李其超,宋 巍,等.基于近岸海面视频的浪高实时检测预测系统[J].计算机技术与发展,2022,32(07):138-143.[doi:10. 3969 / j. issn. 1673-629X. 2022. 07. 024]
　HE Sheng-qi,LI Qi-chao,SONG Wei*,et al.System for Real-time Wave Height Detection and Prediction Based on Near-shore Wave Video[J].,2022,32(07):138-143.[doi:10. 3969 / j. issn. 1673-629X. 2022. 07. 024]
点击复制

基于近岸海面视频的浪高实时检测预测系统()

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
32
期数:
2022年07期
页码:
138-143
栏目:
应用前沿与综合
出版日期:
2022-07-10

文章信息/Info

Title:
System for Real-time Wave Height Detection and Prediction Based on Near-shore Wave Video
文章编号:
1673-629X(2022)07-0138-06
作者:
何盛琪1; 李其超1; 宋 巍1; 王文娟1; 高 松2; 毕 凡2
1. 上海海洋大学 信息学院,上海 201306;
2. 国家海洋局北海预报中心,山东 青岛 266061
Author(s):
HE Sheng-qi1 ; LI Qi-chao1 ; SONG Wei1* ; WANG Wen-juan1 ; GAO Song2 ; BI Fan2
1. School of Information Technology,Shanghai Ocean University,Shanghai 201306,China;
2. North China Sea Marine Forecasting Center of State Oceanic Administration,Qingdao 266061,China
关键词:
浪高检测系统; 浪高预测; 视频信息; 深度学习; 长短期记忆网络
Keywords:
wave height detection system; wave height prediction; video information; deep learning; long short-term memory
分类号:
TP315
DOI:
10. 3969 / j. issn. 1673-629X. 2022. 07. 024
摘要:
针对目前国内近岸浪高监测手段有限、监测频率和精度难以保障的问题,设计和研发了近岸海面视频监控下的浪高实时检测预测系统。 根据业务需求,将系统分为视频传输、视频预处理、浪高检测、浪高预测和浪高可视化五个模块,其中视频传输模块负责所有监控视频在系统中的传输;视频预处理模块负责消除视频中干扰物对浪高检测精度的影响;浪高检测模块使用多层局部感知卷积神经网络( NIN)和支持向量回归( SVR) 对海浪视频进行浪高检测;浪高预测模块使用长短期记忆网络( LSTM)对未来 12 小时和 24 小时的浪高进行预测;浪高可视化模块负责将检测所得浪高值映射为伪彩色,对浪高进行可视化。 该系统支持多观测站点并行检测、站点切换、自动截断保存监控视频和存储浪高数据等功能。 在国家海洋局北海预报中心的应用试验表明,该系统运行稳定,能够较好地协助浪高预报人员的工作。
Abstract:
A real-time detection and prediction system for near-shore wave height based on videos is designed and developed to solve the problems of high operating and maintenance cost of current near-shore wave height detection and difficulty in guaranteeing detection frequency and accuracy. With the B / S architecture,the system is mainly divided into five modules:video transmission module,video preprocessing module,wave height detection module,wave height prediction module and wave height pseudo -color visualization module.The video transmission module is responsible for transcoding the surveillance video into the system,and pushing the real-time monitoring screen of each site to the Nginx server for users to switch and view. The video preprocessing module is responsible for eliminating the influence of interference in the video on the accuracy of wave height detection. Wave height detection module uses convolutional neural network ( CNN) and support vector regression ( SVR) to detect wave heights in ocean wave videos. Wave height prediction module uses long short-term memory neural network ( LSTM) to predict wave heights in the future. Wave height pseudo-color visualization moduleis responsible for converting the detected wave height values into pseudo - color and visualizing the wave height. The system supports parallel detection of multiple observation sites, site switching,automatic cut - off and preservation of surveillance video and storage of wave height data. Applications in the East China Sea Bureau of Ministry of Natural Resources and the North China Sea Bureau of Ministry of Natural Resources show that the system is stable and can better assist the wave height forecasters.
更新日期/Last Update: 2022-07-10