当前位置: X-MOL 学术Water Res. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Lagoon water quality monitoring based on digital image analysis and machine learning estimators
Water Research ( IF 11.4 ) Pub Date : 2020-01-08 , DOI: 10.1016/j.watres.2020.115471
Yuanhong Li , Xiao Wang , Zuoxi Zhao , Sunghwa Han , Zong Liu

Lagoon has been used as one of common to treat agricultural wastewater. However, discharging lagoon water without appropriate treatment could because a critical issue that threaten environment and sustainability. It is necessary to monitor the quality of lagoon water before land application. This paper not only proposes an innovative testing method for sewage, but also combines a machine learning methods and spectral processing techniques to analyze the reflectivity of water chemical composition. The nitrogen, phosphorus, bacteria (total coliform), and total solids (TS), are chosen as main indicators for water quality. It was found that the spectral rate of emission or absorption showed a certain correlation with those contaminants. We use machine learning to train three kinds of estimators to prove those correlation, where their core algorithms are normal equation linear regression (LR), stochastic gradient descent (SGD) and ridge regression (R-PLS). At last, the model dataset is evaluated by weight coefficient, function intercept and mean squared error (MSE). The conclusion shows the reflectivity of TS and spectral reflectance of samples close to linear relationship. The MSE of prediction set and decision coefficient are 0.57 and 0.98, respectively. For bacteria, the MSE of prediction set is 0.63, and coefficient R2 is 0.96. The results from this study could provide a versatile method for remote sensing monitoring of agricultural water pollution and a non-destructive spectroscopy technical reference for agricultural wastewater treatment.



中文翻译:

基于数字图像分析和机器学习估计器的泻湖水质监测

泻湖已被用作处理农业废水的常用方法之一。但是,不经适当处理就排放泻湖水可能是因为威胁环境和可持续性的关键问题。土地施用前有必要监测泻湖水的质量。本文不仅提出了一种创新的污水测试方法,而且还结合了机器学习方法和光谱处理技术来分析水化学成分的反射率。选择氮,磷,细菌(总大肠菌群)和总固体(TS)作为水质的主要指标。发现发射或吸收的光谱速率与那些污染物显示出一定的相关性。我们使用机器学习来训练三种估计量,以证明这些相关性,其中的核心算法是正态方程线性回归(LR),随机梯度下降(SGD)和山脊回归(R-PLS)。最后,通过权重系数,函数截距和均方误差(MSE)评估模型数据集。结论表明TS的反射率和样品的光谱反射率接近线性关系。预测集的MSE和决策系数的MSE分别为0.57和0.98。对于细菌,预测集的MSE为0.63,系数R 预测集的MSE和决策系数的MSE分别为0.57和0.98。对于细菌,预测集的MSE为0.63,系数R 预测集的MSE和决策系数的MSE分别为0.57和0.98。对于细菌,预测集的MSE为0.63,系数R2为0.96。这项研究的结果可以为农业水污染的遥感监测提供一种通用的方法,并为农业废水处理提供无损光谱技术参考。

更新日期:2020-01-09
down
wechat
bug