天津科技 ›› 2023, Vol. 50 ›› Issue (9): 66-72.

• 科学与社会 • 上一篇    下一篇

基于机器学习算法的连云港大浦河水质参数遥感反演

王经顺, 陈莹, 张悦, 徐向凯, 何超, 何苗   

  1. 江苏省环保集团有限公司,江苏省生态环境监测监控有限公司 江苏南京 210005
  • 收稿日期:2023-09-06 出版日期:2023-09-25 发布日期:2023-12-26
  • 基金资助:
    江苏省环保集团科研项目“空天地一体化监测技术在湖泊富营养化监控领域的应用”(JSEP-JC-2022-0166-S-RD)

Remote Sensing Inversion of Water Quality Parameters of Dapu River in Lianyungang Based on Machine Learning Method

WANG Jingshun, CHEN Ying, ZHANG Yue, XU Xiangkai, HE Chao, HE Miao   

  1. Jiangsu Ecological Environment Monitoring and Monitoring Co.,Ltd.,Jiangsu Environmental Protection Group Co.,Ltd.,Nanjing 210005,Jiangsu Province,China
  • Received:2023-09-06 Online:2023-09-25 Published:2023-12-26

摘要: 以连云港市内大浦河、大浦副河部分河段为研究区,根据现场实测数据与无人机高光谱遥感影像结合对水质污染进行研究分析。无人机高光谱数据处理采用多层感知机(MLP)构建水质参数预测反演模型,利用 K-Fold 进行交叉验证,其中氨氮和总磷预测值与实测值的决定系数(R2)高于0.8,平均绝对百分比误差(MAPE)在15%以内。高锰酸盐指数的R2为0.58,原因为其数值范围较小,但总体预测效果理想,MAPE仅为2%左右,反演效果良好。最终通过反演结果对目标河段的污染源现状进行剖析,对主要水环境治理问题的解决提供依据。

关键词: 无人机, 高光谱, 机器学习, 水质参数, 遥感反演

Abstract: This study takes parts of Dapu River and Dapu vice River in Lianyungang City as the study area,and conducts research and analysis on water pollution through field measured data and UAV hyperspectral remote sensing images. Multi- layer perceptron (MLP) is used to construct the prediction and inversion model of water quality parameters for UAV hyperspectral data processing,and K-Fold cross-validation is used. The coefficient of determination (R2) between the predicted values of ammonia nitrogen and total phosphorus and the measured values is higher than 0.8,and the mean absolute percentage error (MAPE) is less than 15%. The R2 of permanganate index is 0.58,because its numerical range is small,but the overall prediction effect is ideal,the MAPE is only about 2%,and the inversion effect is good. Finally,the current situation of pollution sources and the main water environment problems of the target reach are analyzed and discussed through the inversion results.

Key words: UAV, hyperspectral, machine learning, water quality parameters, remote sensing inversion

中图分类号: