Correction And Inversion Of Soil Heavy Metal Content Model Based On CASI/SASI Hyperspectral Data

During the period of rapid economic development in China,sewage irrigation and industrial production have caused serious soil heavy metal pollution,which has caused great harm to humans,animals and plants.Cultivated land heavy metal pollution has become a major ecological issue of human concern.Monitoring of heavy metal pollution in soil is the foundation of soil management and has great practical significance.Hyperspectral remote sensing has the characteristics of rapid,in-situ,real-time,and macro monitoring.It can realize large-scale in-situ monitoring from point to surface,from qualitative analysis to quantitative analysis,and provides a new technical means for soil heavy metal pollution monitoring.In this paper,Xiongan New Area is used as the research area.The whole-region CASI/SASI hyperspectral remote sensing image and the 226 soil samples reflectance data were measured indoor and soil heavy metal content are used as data sources.Use S-G convolution smoothing,envelope removal,reciprocal logarithm,first order,second order differential and other mathematical transformations to transform the measured soil spectrum and soil image spectrum,and establish the functional relationship between the transformed reflectance data and the content of heavy metal elements in the soil,construct a multiple stepwise regression model to realize the inversion of the heavy metal content in the whole region.Finally,the accuracy of the model is checked with the R~2,RMSE and RPD.The main research contents and research results are as follows:(1)Soil sample collection and spectral processing.226 soil samples were collected in the study area,and the indoor spectral curve of soil samples collected by SVC ground spectrometer.Using IBM SPSS21.0 to analyze the correlation between spectral variables and heavy metal content.The pre-treated spectral reflectance data can not only improve the correlation between spectral variables and heavy metal content,but also highlight the masked soil spectral characteristic band.The differential processing technique has the best effect among the soil spectrum pretreatment methods,it can effectively eliminate the interference of the environmental background.(2)Hyperspectral remote sensing image preprocessing.After radiometric calibration,orthorectification,geometric correction,and atmospheric correction of CASI/SASI hyperspectral remote sensing images in the study area,the mask data of dense vegetation,buildings,and water bodies in the image is established,and finally the corresponding soil samples on the image after mask extraction Image spectral data.(3)Establish a multiple stepwise regression model of heavy metal content using the relationship between measured spectral variables and heavy metal content.Through precision evaluation,the best measured spectral variable models of As,Cd,Cu,Pb elements all have the ability to roughly estimate the content of heavy metals.The modeling R~2 of the optimal laboratory spectral model of Cr and Hg elements is above0.5.It is verified that R~2 is 0.41 and 0.43 respectively,and the RPD is lower than 1.4.The model accuracy is low and it is not applicable in the study area.(4)Establish a multiple stepwise regression model of heavy metal content using the relationship between image spectral variables and heavy metal content.Through the accuracy evaluation,The modeling R~2 of Cd,Cr and Cu elements are all above 0.5,the RPD is greater than 1.4,the regression model has better fitting and strong stability;the verification accuracy of As,Hg and Pb element models is lower and the model is not applicable in the study area.(5)Correcting the image spectral model using the measured spectral model can improve the accuracy of soil heavy metal content inversion.Through model correction,the first-order differential spectral model of the As element image verified that R~2was increased from 0.47 to 0.57,and RPD was increased from 1.24 to 1.4;The first-order differential spectral model of the Pb element image verified that R~2 increased from 0.52to 0.62,and RPD increased from 1.28 to 1.61.Select the optimal image spectral regression equation of heavy metal elements,and realize the inversion of the content of each heavy metal element through CASI/SASI hyperspectral remote sensing image.The spatial distribution map of heavy metal content shows that the heavy metal pollution in the cultivated soil around the residential area is relatively serious,which is closely related to human activities and transportation.