Research of Total Phosphorous Detection Combined Multi-Spectra Spectroscopy with Neural Network Algorithm

Abstract

The application of multispectral remote sensing to water quality parameter detection has a good application prospect, but at present, most inversion models based on multispectral water quality parameters use an empirical statistical model, which has poor universality. To solve the above problems, 41 sentinel-2A satellite multi-spectral remote sensing data were collected from Poyang Lake, 35 of which were used as training sets. A prediction model of total phosphorus concentration in water quality based on the neural network algorithm was established. The correlation coefficient of internal verification and testing of the model was above 0.8. The remaining six groups of spectral data were used as external test sets, and the correlation coefficient between predicted value and true value was 0.88, and the root mean square error was 0.048. The experimental results show that the multi-spectral total phosphorus concentration detection model based on neural network algorithm is a feasible water quality remote sensing detection technology