| 引用本文: | 马腾耀,肖鹏峰,张学良,段洪涛,邱银国.基于视频监控图像的环巢湖蓝藻实时动态监测.湖泊科学,2022,34(6):1840-1853. DOI:10.18307/2022.0605 |
| Ma Tengyao,Xiao Pengfeng,Zhang Xueliang,Duan Hongtao,Qiu Yinguo.Real-time monitoring of cyanobacterial blooms dynamics around Lake Chaohu based on video surveillance images. J. Lake Sci.2022,34(6):1840-1853. DOI:10.18307/2022.0605 |
|
| 本文已被：浏览 3197次 下载 1922次 |
码上扫一扫！ |
|
|
| 基于视频监控图像的环巢湖蓝藻实时动态监测 |
|
马腾耀1, 肖鹏峰1, 张学良1, 段洪涛2, 邱银国2
|
|
1.南京大学地理与海洋科学学院, 自然资源部国土卫星遥感应用重点实验室, 江苏省地理信息技术重点实验室, 南京 210023;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
|
|
| 摘要: |
| 蓝藻的防控与治理是湖泊水环境、水生态管理的重要内容，实时获取蓝藻的空间分布信息对于降低蓝藻灾害风险具有重要意义.针对地面调查费时费力、卫星遥感监测粒度较粗且时效性不强等问题，本文提出了一种基于视频监控网络的湖泊蓝藻实时监测技术.基于环巢湖视频监控网络的33个功能摄像机，研究如何从视频图像中实时、准确提取蓝藻的分布信息.为克服不同摄像头的观测角度不一致、光照强度和背景条件不一致等诸多挑战，在视频图像蓝藻表征分析的基础上，通过多尺度深度网络进行图像粗粒度分类，区分蓝藻与浑浊、阴影水体；基于随机森林进行蓝藻精细化识别，克服蓝藻的强异质性.最后以渔政站沿岸水域的日均蓝藻覆盖率和月均蓝藻覆盖率为统计单位，开展了巢湖沿岸蓝藻的动态监测.研究成果可为科学制定蓝藻治理方案提供技术支撑. |
| 关键词: 巢湖 视频图像蓝藻识别 多尺度卷积神经网络 随机森林 蓝藻动态监测 |
| DOI：10.18307/2022.0605 |
| 分类号: |
| 基金项目:安徽省巢湖管理局项目“巢湖蓝藻水华监测预警与模拟分析平台”（2019AMCZ4622）资助. |
|
| Real-time monitoring of cyanobacterial blooms dynamics around Lake Chaohu based on video surveillance images |
|
Ma Tengyao1, Xiao Pengfeng1, Zhang Xueliang1, Duan Hongtao2, Qiu Yinguo2
|
|
1.Jiangsu Provincial Key Laboratory of Geographic Information Science and Technology, Key Laboratory for Land Satellite Remote Sensing Applications of Ministry of Natural Resources, School of Geography and Ocean Science, Nanjing University, Nanjing 210023, P. R. China;2.Key Laboratory of Watershed Geographic Sciences, Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences, Nanjing 210008, P. R. China
|
| Abstract: |
| The prevention and management of cyanobacterial blooms is an important part of lake environment and water ecological management. Real-time acquisition of spatial distribution of cyanobacterial blooms is of great significance for timely salvage and disaster reduction. Aiming at the problems of time-consuming and laborious ground surveys, satellite remote sensing monitoring with a low spatial and temporal resolution, a new method for real-time monitoring of cyanobacterial blooms in lakes using video surveillance network (VSN) was proposed. Based on the 33 cameras of VSN around Lake Chaohu, the study focuses on the real-time and accurate extraction of cyanobacteria distribution information from video images. First, in order to overcome the challenges of different observation angles from different cameras, light intensities and background conditions, the representation of cyanobacterial blooms in video images was analyzed. Then, a multi-scale depth network was used for coarse-grained image classification to distinguish cyanobacteria from turbid and shadowed water; Random forest method was used to finely recognize cyanobacterial blooms to overcome the strong heterogeneity of cyanobacteria. Finally, the distribution information of cyanobacterial blooms was acquired. Based on the average daily and monthly cyanobacteria coverage of the coastal waters of the fishery administration station, monitoring of cyanobacterial blooms dynamics along the Lake Chaohu coast was completed, which can provide technical support for the management of cyanobacterial blooms. |
| Key words: Lake Chaohu cyanobacteria recognition based on video images multi-scale convolutional neural network random forest monitoring of cyanobacterial blooms dynamics |
