A Method of Water COD Retrieval Based on 1D CNN and 2D Gabor Transform for Absorption-Fluorescence Spectra

Area of Science:

• Environmental Science
• Analytical Chemistry
• Spectroscopy

Background:

• Chemical Oxygen Demand (COD) is a key indicator of organic water pollution.
• Accurate and rapid COD detection is crucial for effective environmental protection.
• Existing absorption spectrum methods face challenges with retrieval errors for fluorescent organic matter.

Purpose of the Study:

• To develop a rapid synchronous method for retrieving Chemical Oxygen Demand (COD) using combined absorption and fluorescence spectra.
• To improve the accuracy and reduce errors in COD detection, particularly for fluorescent organic matter in water.
• To introduce an advanced fusion neural network algorithm for enhanced water quality analysis.

Main Methods:

• Development of a fusion neural network algorithm integrating a one-dimensional convolutional neural network (CNN) and 2D Gabor transform.
• Application of the absorption-fluorescence spectrum fusion method to analyze water samples.
• Validation of the method using amino acid aqueous solutions and actual sampled water spectral datasets.

Main Results:

• The absorption-fluorescence COD retrieval method achieved a Relative Root Mean Square Error of Prediction (RRMSEP) of 0.32% in amino acid solutions, an 84% improvement over the single absorption spectrum method.
• COD retrieval accuracy reached 98%, a 15.3% increase compared to the single absorption spectrum method.
• On real water samples, the fusion network demonstrated superior COD measurement accuracy, reducing RRMSEP from 5.09% to 1.15% compared to the absorption spectrum CNN network.

Conclusions:

• The proposed absorption-fluorescence spectrum fusion method significantly enhances the accuracy and reliability of Chemical Oxygen Demand (COD) retrieval in water.
• The fusion neural network algorithm effectively overcomes the limitations of single-spectrum methods, offering a promising tool for environmental monitoring.
• This advanced technique provides a more precise and efficient approach to assessing organic pollution levels in water bodies.