基于视频监控的湖泊滨岸带蓝藻水华实时定量监测方法及应用研究  

Real-time quantitative monitoring method of cyanobacterial blooms in lake riparian zones based on video surveillance and application research

作　　者：刘傅彰 邱银国 张存勇 侯建康 王金迪 刘佳鑫 龙佳莹 马聪 张梦凡 LIU Fuzhang;QIU Yinguo;ZHANG Cunyong;HOU Jiankang;WANG Jindi;LIU Jiaxin;LONG Jiaying;MA Cong;ZHANG Mengfan(School of Marine Technology and Geomatics,Jiangsu Ocean University,Lianyungang 222005,China;Key Laboratory of Watershed Geographic Sciences,Nanjing Institute of Geography and Limnology,Chinese Academy of Sciences,Nanjing 210008,China;School of Surveying and Mapping and Geographical Sciences,Liaoning University of Engineering and Technology,Fuxin 123000,China)

机构地区：[1]江苏海洋大学海洋技术与测绘学院,连云港222005 [2]中国科学院南京地理与湖泊研究所中国科学院流域地理学重点实验室,南京210008 [3]辽宁工程技术大学测绘与地理科学学院,阜新123000

出　　处：《环境工程学报》2024年第2期614-622,共9页Chinese Journal of Environmental Engineering

基　　金：国家自然科学基金资助项目(42101433);江苏省自然科学基金资助项目(BK20201100)。

摘　　要：湖泊作为我国主要饮用水源地,其滨岸带蓝藻水华过度堆积,对用水安全以及生态环境造成严重影响。因此,实时定量监测湖泊滨岸带蓝藻水华,是湖泊蓝藻水华防控的关键举措。基于环湖实时视频监控所捕获的水域图像,采用深度学习方法识别蓝藻水华像素,并根据摄像头成像原理及内外参数准确计算每个蓝藻水华像素对应实际蓝藻水华面积,最后统计分析湖泊滨岸带水域的蓝藻水华总面积。实验结果表明,基于VGG16-UNet模型的蓝藻水华像素识别方法的平均交并比与总体精度分别达到了88.74%与94.10%,优于同类方法;且蓝藻水华面积计算值与实测值具有良好的拟合度(R^(2)=0.97)。该方法能够及时获取湖泊滨岸带蓝藻水华的覆盖面积及其动态变化过程,对于湖泊水环境治理起着重要作用。Excessive accumulation of cyanobacterial blooms in the riparian zones of lakes,which are the main source of drinking water,has a serious impact on water safety as well as on the ecological environment.Therefore,real-time quantitative monitoring of cyanobacterial blooms in lake riparian zones is a key initiative for the prevention and control of cyanobacterial blooms in lakes.Based on the images of waters captured by real-time video surveillance around the lake,a deep learning method was used for cyanobacteria pixel recognition,and the actual cyanobacteria bloom area corresponding to each cyanobacteria pixel was accurately calculated according to the camera imaging principle and internal and external parameters,and finally the total area of cyanobacteria bloom in the waters of the lake’s riparian zone was statistically analyzed.The experimental results showed that the average intersection ratio and overall accuracy of the cyanobacterial pixel identification method based on the VGG16-UNet model reached 88.74% and 94.10%,respectively,which was better than similar methods;and the calculated values of cyanobacterial bloom area had a good fit with the measured values(R^(2)=0.97).This method enabled the timely acquisition of cyanobacterial bloom coverage and facilitates monitoring the dynamic changes along the riparian zones of lakes,playing a pivotal role in the management of lake water environments.

关 键 词：湖泊 富营养化 蓝藻水华 视频监控 定量监测 

分 类 号：X84[环境科学与工程—环境工程]