Inversion of urban river water quality based on machine learning and hyperspectral remote sensing
To address the current lack of research on total phosphorus(TP)and total nitrogen(TN)algorithms in urban rivers,this study focuses on investigating the levels of TP and TN in the Jin'an Rive.Xiadong River will serve as the validation object in this research.The single-band sequential projection algorithm(SPA)and the dual-band Pearson correlation analysis(Pearson)were proposed for independent variable screening.These methods were combined with machine learning algorithms to construct models and visualize water quality inversion based on the measured spectral data of the Jin'an River.The experiment shows that TP,TN and ammonia nitrogen have a strong correlation.When building the inversion model of TP and TN water quality parameters,the TP model with the influence band of ammonia nitrogen added and the SPA+Person+RF algorithm is the best,with R2of 0.92 and ERMS of 0.005 mg·L-1.The TN model with SPA+Pesrson+SVR algorithm is the best,with R2 of 0.90 and ERMS of 0.082 mg·L-1.The optimized algorithm is significantly improved than the traditional algorithm.It is verified that the method is also applicable to the water quality inversion of Xia'dong River,and can be used for urban river water environment monitoring.
water quality inversionunmanned aerial vehicle hyperspectralmachine learningtotal phosphorustotal nitrogenurban river
万方数据
维普
