基于高光谱的小型湖库总氮反演模型研究  

Preliminary Study on Total Nitrogen Inversion Model of Small Lake Based on Hyperspectral Data

作　　者：唐阳 杨远坤 封小倩 谌书[1] 王彬[1] 宋怀庆[1] TANG Yang;YANG Yuankun;FENG Xiaoqian;CHEN Shu;WANG Bin;SONG Huaiqing(School of Environment and Resources,Southwest University of Science and Technology,Mianyang 621010,China)

机构地区：[1]西南科技大学环境与资源学院,四川绵阳621010

出　　处：《水文》2023年第5期33-38,62,共7页Journal of China Hydrology

基　　金：四川省重点研发计划(2020YFS0306,2021YFS0289);四川省重大专项计划(2018SZDZX0026)。

摘　　要：我国小型湖库数量多,分布广,承担的服务功能多样,故构建针对小型湖库的水质及水生态监测体系有重要的科学价值。利用实时获取的高光谱数据,运用归一化、波段比值和波段差值解析敏感波段,建立多元线性回归模型和BP神经网络模型反演西南科技大学中心湖(小型人工湖)总氮浓度,并分析模型的精度与适用性。结果显示,采用波段比值运算解析获取敏感波段,并将其作为输入层的神经网络模型拟合精度最高(均方根误差为0.05mg/L,平均绝对误差为0.03mg/L,平均相对误差为4.7%)。结果表明,高光谱遥感的连续光谱结合BP神经网络的非线性拟合能力可较好地反演总氮浓度。研究结果有助于为小型湖库高光谱水质反演及监测预警提供理论和技术参考。There are many small lakes and reservoirs in China,which are widely distributed and have various service functions.Thus,it is of great importance to establish a monitoring system for water quality and water ecology of them.In this study,the hy⁃perspectral data obtained in real time was used to analyze sensitive bands by band normalization,band ratio and band difference.Then the multiple linear regression model and BP neural network model were established to retrieve the total nitrogen(TN)con⁃centration in the central lake(a small artificial lake),which is located at Southwest University of Science and Technology.Finally the accuracy and applicability of the models were analyzed.The results show that the neural network using sensitive band after band ratio operation as the input layer has the highest fitting accuracy(root mean square error is 0.05 mg/L,the average absolute error is 0.03 mg/L and the average relative error is 4.7%).These indicate that the continuous spectrum of hyperspectral remote sensing combined with the nonlinear fitting ability of BP neural network could better retrieve the concentration of TN.The results could provide theoretical and technical reference for hyperspectral water quality inversion,monitoring and early warning in small lakes and reservoirs.

关 键 词：小型湖库 总氮浓度反演 高光谱 神经网络模型 敏感波段解析 

分 类 号：P342[天文地球—水文科学]