|
| 南四湖流域常规水污染物多源遥感监测系统设计 |
| Design of Multi source Remote Sensing Monitoring System for Conventional Water Pollutants in Nansi Lake Basin |
| |
| DOI： |
| 中文关键词: 水污染物 遥感监测 自适应反演 机器学习 南四湖流域 |
| 英文关键词: Water pollutants Remote sensing monitoring Adaptive inversion Machine learning Nansi Lake basin |
| 基金项目:国家自然科学面上基金资助项目（42276185）；山东省本科高校教学改革研究基金资助项目（M2022040）；西海岸新区高校校长专项基金资助项目（GXXZJJ202303） |
|
| 摘要点击次数: 103 |
| 全文下载次数: 55 |
| 中文摘要: |
| 提出一种多源遥感监测方法，用于南四湖流域常规水污染物监测系统设计。超分辨率处理后的Sentinel 2 L2A卫星多光谱数据可实现流域水质的大范围监测，利用无人机多光谱/高光谱数据可实现入湖河流水质的精细化监测，地物光谱仪高光谱数据能够实现湖、河岸边带水体的水质反演精度修正，基于多源遥感影像光谱特征与非光敏水质参数的相关性分析，建立光谱特征数据库。相比半经验法，结合机器学习法后，总氮、总磷反演的平均误差分别下降2284%、3919%。通过增强数据集、添加正则化项、集成学习、泛函分析等正则化技术，能够进一步提高反演模型对不同区域、不同季节、不同水质类别水体的自适应能力。 |
| 英文摘要: |
| A multi source remote sensing monitoring method was established for the design of conventional water pollutant monitoring system in Nansi Lake basin. The super resolution multispectral data from Sentinel 2 L2A satellite could be used for large scale monitoring of water quality in the basin. The multispectral / hyperspectral data by UAV was used for fine monitoring of the quality of the water entering lakes and rivers. The inversion accuracy of water quality in the lake and riparian zones was corrected by hyperspectral data measured by field spectrometer. A spectral feature database was established based on correlation analysis between spectral features of multi source remote sensing images and non photosensitive water quality parameters. Compared with semi empirical method， the average errors of TN and TP inversion by this method combined with machine learning method decreased by 2284% and 39.19% respectively. Through the regularization technologies of enhancing dataset， adding regularization terms， ensemble learning and functional analysis， the adaptive ability of the inversion model to different regions， different seasons and different water quality categories could be further improved. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
