Study On Hyperspectral Quantitative Inversion About PH Of Water And Zn Content Of Soil In Mining Areas

Currently,water and soil resources have been severely polluted by extensive and unreasonable mining.For example,acidic mine drainage can cause severe acidification of surrounding water;coal mine dust can lead to heavy metal pollution in surrounding soil.Therefore,it is of great importance to monitor and govern the mining actions,thus protecting our living environment.With the unique advantages of large scale and rapid environment monitoring,hyperspectral data can obtain the continuous spectral curves of the objects and improve the accuracy of quantitative analysis of remote sensing.In this study,Dexing Copper Mine and Shengli Coal Mine are taken as the research earas.Spectrums of water and soil in two mining areas are obtained based on remote-sensing images of unmanned aerial vehicle(UAV)and laboratory spectrometer,respectively.Get the useful Spectrums by de-noising water spectrums and removing the water vapor absorption band of soil spectrums.After analyzing spectral response features and screening feature factors,the inversion models are constructed,and the quantitative inversion of water pH value and soil Zn content in the mines are finally realized.The main research contents and findings are as follows:(1)On the basis of the two-band ratio,a reflectance transformation method of multi-band ratio is proposed,which is used to excavate the spectral characteristics.The findings show that the spectral reflectances are strongly correlated with pH and Zn content after the ratio transformation,where the multi-band combination ratio proposed in this paper has the highest correlation.(2)Spectral response characteristics of water pH and soil Zn are analyzed quantitatively.Three groups of pH value characteristic factors and seven groups of Zn content characteristic factors are screened in light of the correlation coefficient for quantitative inversion.(3)The models of multiple regression,stepwise regression and support vector machine regression(SVR)for pH value and Zn content are constructed,and the comparative analysis and evaluation are performed.Multiple regression model has the lowest accuracy;stepwise regression is slightly higher than multiple regression in the accuracy because it could screen optimal independent variables;SVR model is identified as the best inversion model with the best inversion effect and the highest accuracy.The nonlinear model has proven to be universal in hyperspectral quantitative inversion.