Hyperspectral Inversion Method Of Heavy Metal Content In Subsidence Waters Of Coal Mines

The subsidence water area of the mining area is generally a closed system and does not circulate with the outside world.It is not easy to remove harmful substances.It is greatly affected by the coal dust and the surrounding environment,and its main water source carries peripheral pollutants,which in turn affects industrial and agricultural production.Hyperspectral remote sensing technology has the obvious features that are difficult to achieve with conventional water monitoring methods and can perform highly accurate water quality detection.This paper takes the subsidence area of Panji Chuang Da Ecological Park in Huainan City as the research object,and collects the spectral data of the water sampling points through the field spectroscopy instrument.At the same time,sampling in the field and performing laboratory analysis,six heavy metal elements concentration(Cu,Pb,Zn,Zn,Cd,Cr)are measured.Based on the correlation statistics between the remote sensing data and the measured data of the sampling points,the hyperspectral inversion model for the metal content of water body weight was selected by combining the characteristic bands and bands,and the determination coefficients and the root mean square errors of the inversion models were used as the evaluation criteria.Select the best inversion model for each heavy metal element and its accuracy.Experimental results show:(1)The order of the coefficient of variation of the metal elements in the study area is Cu>Zn>As>Pb>Cr>Cd.Six kinds of sample water Heavy metals content of skewness and kurtosis are not equal to 0;the correlation coefficient between Cu and As,Pb,and Cr was relatively high,and their distribution was affected by some common factors,with similar pollution sources and pollution transfer processes.(2)Based on different spectral forms of spectral features and dual-band ratios,the heavy metal contents in subsidence waters were fitted with different models(linear,logarithmic,quadratic,power,exponent).The best inversion models for six heavy metals were obtained.The single-band inversion model with water metal content in the water has the highest prediction accuracy for Cr,spectral inversion model in the form of second-order differential transformation works best,and the quadratic model for the prediction accuracy of various metals.The inversion model based on the dual-band ratio does not always improve the prediction accuracy of the six heavy metal contents compared to the single-band inversion model.For some heavy metals,the accuracy is even greatly reduced.(3)The stepwise multivariate regression model of various metals obtained by spectral second-order differential transformation has good stability and high precision.The fitting degree between predicted and measured values of water body weight metal content is higher,and the estimation effect of each regression equation is better than the other three methods,which can provide basis and support for high-spectrum rapid estimation of metal content of water body weight in subsidence waters of the area.The accuracy of the model for inverting heavy metal content based on the BNA band depth feature index is higher than that of other band depth feature index inversion models.The deterministic coefficients of the various band depth models for heavy metal Cu are higher than 0.6,which has a certain reference significance.