Quantitative Inversion Of Heavy Metal Chromium In Soil Based On Multi-source Hyperspectral Data

Land is an important material resource for people to survive.It nourishes everything,but due to heavy industry,mining,sewage irrigation,etc.,the problem of land pollution is becoming more and more serious.Especially the soil heavy metal pollution,it has a certain concealment,it is difficult to be discovered,and it is extremely harmful,so the country has begun to increase efforts to rectify the soil environment.It is common to survey the status of heavy metal pollution in the soil,using traditional geochemical methods,which require a lot of manpower,material resources and financial resources.With the development of hyperspectral remote sensing technology,the inversion map of soil heavy metal content can be realized relatively quickly through soil spectral information and fewer sample points,providing auxiliary information for relevant departments and dynamic monitoring.This paper takes part of cultivated land in Jinzhou City of Liaoning Province as the research object,and explores the spectral characteristics and inversion model of heavy metal chromium Cr in soil by hyperspectral remote sensing technology.Finally,based on airborne CASI,SASI data and onboard HJ-1A HSI data to map.Spectral curves and geochemical data were detected using field samples.Spectral preprocessing was performed using continuum removal,first-order differential,multi-scattering correction,and reciprocal logarithm.The comparison analysis showed that continuum removal,multi-scatter correction,The first-order differential can better highlight the correlation between the spectrum and the heavy metal Cr content.There is no significant correlation between the reciprocal logarithm,the original spectrum and the heavy metal Cr content.The selected characteristic bands were modeled and predicted using multivariate stepwise regression,partial least squares method and neural network respectively.The accuracy of the three models was gradually improved.The model with the highest accuracy of CASI,SASI was neural network,and the R~2 was0.78.The RMSE is 23.93,and the MRE is 0.31.The model with the highest accuracy of HJ-1A HSI is a neural network,and the R~2 is 0.71,the RMSE is 27.91,and the MRE is 0.35.Finally,the best precision model was used to invert the Cr content of heavy metals in the study area by airborne CASI,SASI and spaceborne HJ-1A HSI hyperspectral data.The conclusions are more consistent.The Cr pollution in the study area is relatively small,indicating that the local government has achieved certain success after government remediation measures.The paper combines hyperspectral techniques with statistical methods to explore the spectral characteristics of heavy metal Cr in the visible-near-infrared range of soil,and uses a variety of model methods for comparative analysis to provide a reference for related research.