Research On Water Quality Parameter Monitoring Algorithm Based On Remote Sensing Image

Due to the influence of human production and life,the water in China is polluted to different degrees.It is difficult to obtain the overall water quality by the traditional sampling method.By monitoring water quality parameters with remote sensing images,the monitoring scope of water area is more macroscopic.p H,dissolved oxygen?DO?,chemical oxygen demand?COD?and NH₃-H content are important basic parameters for evaluating water quality in China.However,the current monitoring algorithm for these four types of water quality parameters is not mature enough,which is a hot topic in current research.Based on remote sensing image,this dissertation studies the monitoring algorithm of these four kinds of water quality parameters.To solve the problem that the above four types of water quality parameters are not highly correlated with the spectral values of each band of landsat-8 remote sensing image,it is difficult to establish the mapping relationship between water quality parameters and a single band,and the accuracy of the traditional multiple regression method is low,this dissertation proposes the BP neural network algorithm applicable to the monitoring of water quality parameters.Based on the advantages of BP neural network with strong nonlinear mapping ability and self-learning and self-adaptive ability,the inversion model of water quality parameters based on spectral values of multiple bands of remote sensing images is established.The experimental results show that the relative root-mean-square errors between the predicted value and the measured value of p H,DO,COD and NH₃-H by BP neural network model are 0.0808,0.3417,0.1970 and 0.3242,respectively.The relative error of p H and COD is small,and the model can reflect the water quality parameters well.The relative error of the test data of DO and nh3-h content is large,but the goodness of fit of the two inversion models of water quality parameters is acceptable.Considering that water quality parameters are time series data with trend,seasonality and periodicity,this dissertation proposes an improved model LSTM using recurrent neural network to monitor water quality parameters.The water quality parameters of the previous period are introduced in the input layer of the algorithm by using the advantage of the recurrent neural network to process the time sequence information.The experimental results show that the inversion results of water quality parameters have been significantly improved,and the relative root-mean-square errors between the predicted values of p H,DO,COD and NH₃-H content and the measured values are 0.0674,0.2099,0.1241 and 0.2519,respectively.The water quality parameters obtained by LSTM inversion based on remote sensing image basically meet the requirements of actual water quality monitoring.The research results provide a broader perspective for future research on water quality parameter monitoring based on remote sensing images.