温排水自动提取方法研究  

Research on automatic extraction method of thermal discharge

作　　者：朱秀芳 李原 ZHU Xiu-fang;LI Yuan(State Key Laboratory of Remote Sensing Science,Beijing Normal University,Beijing 100875,China;Institute of Remote Sensing Science and Engineering,Faculty of Geographical Science,Beijing Normal University,Beijing 100875,China;The Inner Mongolia Key Laboratory of River and Lake Ecology,Hohhot 010021,China;School of Ecology and Environment,Inner Mongolia University,Hohhot 010021,China)

机构地区：[1]北京师范大学,遥感科学国家重点实验室,北京100875 [2]北京师范大学地理科学学部,遥感科学与工程研究院,北京100875 [3]内蒙古自治区河流与湖泊生态重点实验室,内蒙古呼和浩特010021 [4]内蒙古大学生态与环境学院,内蒙古呼和浩特010021

出　　处：《中国环境科学》2023年第11期6115-6122,共8页China Environmental Science

基　　金：国家自然科学基金资助项目(41292583)。

摘　　要：为解决现有的遥感温排水检测的算法自动化程度低、时效性差的问题,提出了一种基于单时相热红外波段的温排水自动化提取方法(IForest-SVM),并以福建省福清核电站为例,使用2018年10月至2019年1月内的4期Landsat-8遥感影像对所提方法进行了测试与验证.该方法首先利用孤立森林进行温排水和正常水体样本的自动提取,接着通过温排水和排水口的空间邻接关系纯化温排水样本,进而利用支持向量机监督分类提取温排水像元,最后通过温排水和排水口的空间邻接关系剔除误判像元得到最终的温排水空间分布.测试结果显示本研究所提出的方法在2018年11月15日、2018年12月1日和2019年1月18日三期图像上检测得到的温排水的用户精度和生产者精度的平均值分别为89.69%、94.97%和90.04%.在2018年10月30日无温排水的影像上没有出现误检,有效的避免了“虚警”的出现.IForest-SVM只需要输入遥感影像的热红外波段,无需额外代入其他参数,具有可移植性好、普适性强、自动化程度高的优点,在温排水的实时监测和快速发现业务中有很好的应用前景.To solve the problems of low automation and poor timeliness of existing algorithms for remote sensing thermal discharge detection,this study proposes an automated extraction method for thermal discharge based on the single-temporal thermal infrared band,called IForest-SVM.The Fuqing Nuclear Power Plant in Fujian Province was taken as an example to test and verify the proposed method using four Landsat-8 remote sensing images from October 2018 to January 2019.IForest-SVM first uses isolation forest to automatically extract samples of thermal discharge and normal water bodies,then purifies the thermal discharge samples by their spatial adjacency relationship with discharge outlets.After that,a support vector machine is used to extract thermal discharge pixels,and misclassified pixels are removed based on their spatial adjacency relationship with discharge outlets to obtain the final spatial distribution of thermal discharge.The test results show that the average values of user accuracy and producer accuracy of thermal discharge detected by the proposed method on three images on November 15,2018,December 1,2018,and January 18,2019 were 89.69%,94.97%,and 90.04%,respectively.No false alarms occurred on the image without thermal discharge on October 30,2018,effectively avoiding false detections.IForest-SVM only requires input of the thermal infrared band of remote sensing images without additional parameters,and has the advantages of good portability,strong applicability,and high automation.It has a good application potential in the real-time monitoring and rapid detection of thermal discharge.

关 键 词：孤立森林 异常检测 支持向量机 邻接关系 温排水 

分 类 号：X87[环境科学与工程—环境工程]