基于HJ-2A/B卫星高光谱数据的湖泊颗粒态磷遥感估算  

Estimation of Lake Particulate Phosphorus by Remote Sensing Based on Hyperspectral Data from the HJ-2A/B Satellite

作　　者：万能胜 程宏伟 余寒明 程佳美 唐晓先 杨富宝 熊竹阳 齐鹏云 潘邦龙 WAN Nengsheng;CHENG Hongwei;YU Hanming;CHENG Jiamei;TANG Xiaoxian;YANG Fubao;XIONG Zhuyang;QI Pengyun;PAN Banglong(Lake Ecological Environment Research Institute,Chaohu Lake Administration Bureau,Hefei 230601,China;Environment and Energy Engineering Department,Anhui University of Architecture,Hefei 230601,China;Anhui Water Resources and hydropower Survey design Research Institute Co.LTD,Hefei 230601,China;Anhui Fenghuang Neck Irrigation and Drainage Station Management Office,Wuhu 238341,China)

机构地区：[1]安徽省巢湖管理局湖泊生态环境研究院,安徽合肥230601 [2]安徽建筑大学环境与能源工程学院,安徽合肥230601 [3]安徽省水利水电勘测设计研究总院有限公司,安徽合肥230601 [4]安徽省凤凰颈排灌站管理处,安徽芜湖238341

出　　处：《洛阳理工学院学报（自然科学版）》2024年第4期1-8,92,共9页Journal of Luoyang Institute of Science and Technology：Natural Science Edition

基　　金：国家自然科学基金项目(42277075);安徽省自然科学基金水科学联合基金项目(2208085US14);安徽省高校自然科学基金项目(2023AH050187).

摘　　要：以巢湖为研究对象,利用HJ-2A/B卫星HSI高光谱遥感数据,结合地面实测样点数据,采用Pearson相关分析法和CARS算法筛选颗粒态光谱敏感波段,结合随机森林(RF)、支持向量机(SVM)、极限学习机(ELM)、卷积神经网络(CNN)等机器学习算法构建湖泊颗粒态磷遥感估算模型,并进行反演。Pearson-CNN模型具有较好的预测能力,R^(2)、RMSE、RPD分别为0.765、32.49μg/L、1.97。这表明该模型能够快速捕获颗粒态磷光谱特征,具有较强的非线性学习能力,有利于提高颗粒态磷遥感模型的估计精度,为湖泊磷源监测与评估提供了良好的支持。Taking Chaohu Lake as the research area,using HJ-2A/B satellite HSI hyperspectral remote sensing data,combined with ground measured sample data,Pearson correlation analysis and CARS algorithm were used to screen the sensitive spectral bands of particle state.Combined with some machine learning algorithms,such as random forest(RF),support vector machine(SVM),extreme learning machine(ELM),convolutional neural network(CNN),the remote sensing estimation model of lake particulate phosphorus was constructed and inversion was performed.The results showed that the Pearson-CNN model had good predictive ability,with R^(2),RMSE and RPD of 0.765,32.49μg/L and 1.97 respectively.This shows that the model can quickly capture the spectral characteristics of particulate phosphorus and has strong nonlinear learning ability,conducive to improving the estimation accuracy of the remote sensing model of particulate phosphorus,providing good support for monitoring and evaluation of lake phosphorus sources.

关 键 词：颗粒态磷 环境二号卫星 高光谱遥感 机器学习 

分 类 号：TP791[自动化与计算机技术—检测技术与自动化装置]