摘要
以河北省廊坊市为研究区,根据黑臭水体光谱性质,基于CART模型提出一种改进的模糊决策树黑臭水体遥感识别方法。研究数据表明:黑臭水体与一般水体在可见光及近红外波段存在光谱差异,R_(rs)(R)-R_(rs)(B)、R_(rs)(R)-R_(rs)(G)以及R_(rs)(B)+R_(rs)(G)+R_(rs)(R)可以较好地区分两类水体;根据叶子节点隶属度进行黑臭水体提取(黑臭水体<0.5,一般水体>0.5),其中隶属度为1或0的节点定义为置信区,其余节点为模糊区;提取结果总体精度达到84.78%,其中置信区为92.85%,模糊区为72.23%。该方法在实现高精度提取黑臭水体的同时,通过定义置信区和模糊区可有效降低人工核查,实现更为高效的工程应用。
The monitoring and treatment of urban black and odorous water body is a hot issue in the field of water pollution prevention and control.Remote sensing technology has become an important means of black and odorous water body monitoring in recent years because of its macro,efficient and large-scale monitoring.Taking Langfang city,Hebei province as the research area,combined with sentinel-2 A image data,by analyzing the spectral properties of black and odorous water,an improved multi feature fuzzy decision tree classification algorithm is proposed based on cart model algorithm.The results show that:(1) there are obvious spectral differences between black and odorous water bodies and general water bodies.The reflectance difference between green band and blue band,the sum of reflectance between blue band and green band and red band,and the sum of reflectance in visible light band can be well divided into two types of water bodies;(2) by calculating the membership degree of the leaf nodes of the cart decision tree,the black and odorous water body can be extracted.The nodes with membership degree of 1 or 0 are defined as the confidence area,and the other nodes are fuzzy areas.Finally,the results with membership degree less than 0.5 are extracted into the general odorous water body,and those greater than 0.5 are the black and odorous water body;(3) through the verification set test,the overall accuracy of cart decision tree algorithm is 84.78%,the classification accuracy of confidence area is 92.85%,and the node classification accuracy of fuzzy area is 72.23%.While realizing the high-precision extraction of black and odorous water body,the proposed method can further reduce the workload of field verification and realize more efficient engineering application by defining confidence area and fuzzy area.
作者
董旭鑫
赵起超
李家国
李国洪
金永涛
DONG Xuxin;ZHAO Qichao;LI Jiaguo;LI Guohong;JIN Yongtao(School of Electronic and Communication Engineering,Hebei University of Technology,Tianjin 300000,China;Hebei Province Aerospace Remote Sensing Information Processing and Application Collaborative Innovation Center,Langfang,Hebei 065000,China;North China Institute of Aerospace Engineering,School of Remote Sensing Information Engineering,Langfang,Hebei 065000,China;Aerospace Information Research Institute,Chinese Academy of Sciences,Beijing 100101,China)
出处
《遥感信息》
CSCD
北大核心
2022年第5期63-69,共7页
Remote Sensing Information
基金
国防基础科研计划(JCKY2019407D004)
河北省自然科学基金青年基金项目(D2020409005)
河北省高等学校科学技术研究项目(ZD2019138)。
关键词
遥感应用
黑臭水体
廊坊市
多特征
分类
CART算法
remote sensing application
black and odorous water body
Langfang city
multiple-feature
classification
CART algorithm