Quantitative Sudyon Watert Quality Idex Bn Ased On Hyperspectral Technology

Water quality detection is the premise and foundation of water protection.Biochemical oxygen Gen demand(BOD),chemical oxygen Gen demand(COD),nitrite can reflect the degree of water pollution.The research on the rapid detection of the above parameters can provide effective data support for water resource protection.The quantitative research of water quality index based on hyperspectral technology can realize the real-time detection of water quality quickly,contactless and large area.It is of great scientific significance and practical value to provide theoretical basis for the current water pollution control problem in China.This paper starts with the research on the common methods of water quality monitoring,and investigates the current situation and basic principles of hyperspectral analysis.The water samples are configured,the spectral data of samples are collected,and the data is filtered and preprocessed.The partial least square regression model,BP neural network model and Elman neural network model are studied.The original data and pretreatment data are used to establish the quantitative estimation models of COD,BOD and nitrite.The advantages and disadvantages of the model are analyzed.The influence of different pretreatment methods on the accuracy of the model is analyzed.The characteristic band is extracted by partial least square regression coefficient method,and the BP neural network model and Elman neural network model are optimized.Then,the feature bands under different pretreatment are combined to establish a quantitative model of multi feature fusion hyperspectral water quality index.The experimental results show that the partial least square regression method,BP neural network and Elman neural network can establish a good accuracy quantitative water quality index estimation model;different pretreatment methods can improve the prediction effect of the model;the feature band extraction method can reduce the interference of useless information,improve the correlation between spectral data and index concentration;return partial least square back The combination of regression method and neural network model can improve the performance of quantitative estimation model of water body index.Among them,the multi feature fusion pls BP neural network model has the best prediction effect.Through the test of the simulation software of quantitative analysis of water quality index,the technical standard of the national water quality prediction instrument can be met.It can meet the national requirements for online water quality prediction instrument It is used in water quality prediction instrument.In this paper,a simulation software of hyperspectral quantitative analysis of water quality index is designed by Matlab GUI.The process of the model is visualized to show the prediction results and accuracy of the model,and the quantitative analysis of water quality indexes can be realized.The simulation software has perfect function and convenient operation,and has completed the expected function requirements.