Inversion Of Chlorophyll Concentration In Lake Taihu Based On BP Neural Network

Water,one of the most important ecological elements in the natural environment,is the source of life,especially important for the study and monitoring of water quality.As the most important freshwater resource on the earth,lakes play a major role in maintaining regional ecological balance and maintaining biodiversity.With the rapid development of society and economy,the lake water quality in our country begins to deteriorate continuously,most of which are in the state of eutrophication,Lake Taihu is a typical representative of eutrophication lakes in China.The eutrophication of the water body leads to the gradual deterioration of the ecological environment.Therefore,a scientific method is urgently needed to monitor the pollution status and the changing trend of the Lake Taihu.Chlorophyll-a is an important indicator of water quality,which can be used to evaluate the degree of eutrophication and water pollution.BP neural network has the ability to realize complex nonlinear mapping,which is suitable for prediction and classification.Based on BP neural network,this paper establishes an inversion model suitable for the concentration of chlorophyll-a in Lake Taihu.In this paper,BP neural network is used to study the inversion algorithm of chlorophyll concentration in Taihu Lake.The main contents include:(1)Data preprocessing.Firstly,the measured data of chlorophyll-a concentration at the sampling point in Lake Taihu and the corresponding MERIS satellite remote sensing data were obtained.The satellite remote sensing data were preprocessed,including radiation correction,geometric correction,atmospheric correction and image cutting.Then the radiation value of the input data is analyzed,and 8 more sensitive band radiation values are selected as input parameters.(2)Constructing the model of chlorophyll concentration in Lake Taihu based on BP neural network.The design of BP neural network mainly includes the number of network layers,the number of input layer nodes,the number of hidden layer nodes,the number of output layer nodes and transfer function,the training method and the setting of training rate.After training,adjust the parameters,the final established model whose average relative error is 24.754%,in which input layers contains 8 nodes,the hidden layers contains 10 nodes and output layer nodes contains 10 nodes.(3)BP neural network optimization.In order to overcome the limitation that BP neural network is easy to fall into the local minimum,genetic algorithm is used to optimize the constructed BP neural network.Genetic algorithm optimizes the weights and thresholds of BP neural network through selection,crossover and mutation operations.The improved BP neural network using genetic algorithm is better in accuracy and convergence speed,and the average relative error is reduced to 6.26%,which has a good inversion effect.In this paper,the inverse model of chlorophyll concentration based on BP neural network overcomes the limitations of traditional empirical model in dealing with nonlinear problems,and the genetic algorithm is used to optimize the initial parameters of BP neural network.The optimized model has a better fit effect.The study results provide theoretical support for the research of water color remote sensing based on BP neural network.However,it is necessary to further improve the atmospheric correction of MERIS images,the preprocessing of remote sensing data,and the optimization of network parameters of genetic algorithms,so as to improve the output accuracy of the network.