AUTHOR=Zhang Hongming , Zhou Xiang , Tao Zui , Lv Tingting , Wang Jin 

TITLE=Deep learning–based turbidity compensation for ultraviolet-visible spectrum correction in monitoring water parameters

JOURNAL=Frontiers in Environmental Science

VOLUME=Volume 10 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/environmental-science/articles/10.3389/fenvs.2022.986913

DOI=10.3389/fenvs.2022.986913

ISSN=2296-665X

ABSTRACT=Ultraviolet-Visible spectroscopy was an effective tool for qualitative analysis and quantitative detection of water parameter without reagent. Suspended particles in water cause turbidity interference with ultraviolet-visible spectrum that will finally affect the accuracy of water parameter calculation. This paper proposed a deep learning method to compensate turbidity interference and obtain water parameters with Partial Least Squares Regression method. Compared with orthogonal signal correction (OSC), Extended Multiplicative Signal Correction (EMSC) method, deep learning method specifically a 1-dimensional U-shape neural network (1D U-Net) is accurate and the first time to take turbidity compensation in real river water of agricultural catchment. After turbidity compensation, R2 between predicted value and true value increased from 0.918 to 0.965 and RMSE value decreased from 0.526 to 0.343 mg. The experimental results showed that1D U-Net is suitable for turbidity compensation and obtain accurate result.