农村水电站生态流量泄放视频智能监管系统研发  

Research and Development of an Intelligent Video Monitoring System for Ecological Flow Discharge in Rural Hydropower Stations

作　　者：廖佳庆 刘锋 蒋元中 LIAO Jia-qing;LIU Feng;JIANG Yuan-zhong(Hangzhou Dingchuan Information Technology Co.,Ltd.,Hangzhou 310000,Zhejiang Province,China;Zhejiang Institute of Hydraulics and Estuary(Zhejiang Institute of Marine Planning and Design),Hangzhou 310000,Zhejiang Province,China;Zhejiang Water Conservancy Disaster Prevention and Reduction Quality Inspection Station,Hangzhou 310000,Zhejiang Province,China)

机构地区：[1]杭州定川信息技术有限公司,浙江杭州310000 [2]浙江省水利河口研究院(浙江省海洋规划设计研究院),浙江杭州310000 [3]浙江省水利防灾减灾重点实验室,浙江杭州310000

出　　处：《中国农村水利水电》2024年第2期160-164,171,共6页China Rural Water and Hydropower

基　　金：浙江省水利厅科技计划项目(RC2155)。

摘　　要：我国农村水电站站址分散,泄放设施形式各异造成生态流量泄放监测成本过高,同时河流断面不规则性、水文情势随机性等也增加了流量监测的难度导致安装监测设备不能大量推广。基于现状,通过非接触的视频识别是未来一个有效的监管途径,有利于监管工作的推广,通过对大量的生态流量泄放视频进行研究,研发了视频智能识别算法,通过主算法和子模型两种识别方式相结合实现了对生态流量是否泄流的识别,既保障了识别通用性、易用性,又保障了识别结果的精度。同时研发了水利视频AI通用组件系统整合了视频智能识别算法和识别业务流程,实现了对499个农村水电站生态流量是否泄放的智能监管。通过对5个月的所有识别结果进行人工审核统计,剔除视频或图像模糊的情况,平均识别准确率97.8%,应用情况良好,具备推广价值。The fact that rural hydropower stations are scattered in sites and discharge facilities differ widely in their forms causes the high costs of ecological flow discharge monitoring in China.Meanwhile,the irregularity of river sections and the randomness of hydrological situations also increase the difficulty of flow monitoring,impeding the extensive application of regular monitoring equipments.Under such circumstances,non-contact video recognition is an effective alternative.Based on the data extracted from a large number of ecological flow discharge videos,the authors develop a video recognition intelligent algorithm,which can be used to identify whether the ecological flow has been discharge or not.By combining a main algorithm and a sub-model,this system ensures not only the universality and ease of use of the identification,but also the accuracy of the identification results.The authors also develop a water conservancy video AI general component system,which integrates the video recognition intelligent system and the identification business process,and has been put into use in 499 rural hydropower stations through a period of five months.The effectiveness and promotion value of this system is testified by the statistics of all recognition results:after ruling out the cases where the videos or images are blurred,the average recognition accuracy under normal conditions is 97.8%.

关 键 词：视频识别 生态流量 图像识别 水利监管 

分 类 号：TV93[水利工程—水利水电工程] TP39[自动化与计算机技术—计算机应用技术] TP31[自动化与计算机技术—计算机科学与技术]