Soil Feature Parameter Inversion Modeling Based On Hyperspectral Data

With the rapid development of modern industry and technology,the problem of soil pollution has become more and more serious,which has brought long-term serious threats to people’s health.Due to the non biodegradability of heavy metals,the monitoring and diffusion prevention of soil contaminated with heavy metals has become an important task in the prevention and control of heavy metal pollution.The traditional chemical analysis method for soil heavy metal detection has a low cost-effectiveness ratio and a long time-consuming process,and it is impossible to work quickly in a large area.Hyperspectral remote sensing,a new modern detection technology,can quickly obtain a large amount of clear surface information including material composition.Therefore,it can be used as a technical means for rapid detection of large areas of heavy metals in soil.In this paper,hyperspectral characteristic parameters and soil heavy metal content inversion were studied,and the spectral interpolation surface was established.Based on this,a technical method system for rapid detection of soil characteristic parameters based on hyperspectral data is constructed,and this model is applied to the northeastern sulfonate.In the typical area of the district.The main research contents and conclusions of this paper are as follows:1)Soil geochemical data denoising based on the total variation model.Geochemical data denoising has always been an important part of geochemical data analysis.It is generally believed that the actual measured geochemical data contains noise,and the noise can be removed to obtain true geochemical data.Firstly,the geochemical data is regarded as a two-dimensional image.Secondly,the total variation model used in image denoising is introduced into the geochemical data denoising to construct an overall variational noise reduction model for geochemistry.Soil geochemical data denoising based on the total variation model can predict errors and shorten the length of the distribution of element content while maintaining the overall shape of geochemical data.2)Inversion model of heavy metal content in soil based on spectral features.In this paper,we extract 12 kinds of spectral feature data including original spectral data through spectral differential variation,inverse logarithmic transformation,envelope removal method and various combinations of these methods.The approximate scope of the spectral characteristics are delineated by geometric analysis of spectral curves.The approximate range of the six characteristic bands of each heavy metal element and each spectral data was extracted by correlation analysis of various heavy metal contents with corresponding spectral characteristic data.A coupled model of each heavy metal element was established and elected by stepwise regression analysis.We shows that the coupled model established in this paper is reliable for the prediction of heavy metals in the study area via computation set data validation and non-computation set data validation.We shows that the coupled model established in this paper is reliable for the prediction of heavy metals in the study area via computation set data validation and non-computation set data validation;that it can quickly detect the heavy metal content in the study area by using the model established.Therefore,it also indicates that the technique of soil heavy metal content inversion model based on hyperspectral data is feasible.3)A technical system for rapid monitoring of heavy metal pollution based on hyperspectral technology.We adopts the layout idea from coarse to fine,and uses multi-spectral high-resolution remote sensing image to extract the texture features,color features,Iron stain abnormality and other information related to sulfur slag,and to quickly delineate the spatial distribution of sulfur slag.The spatial interpolation modeling is optimized via comparing the global Kriging interpolation,modified Kriging interpolation and fractal interpolation.The spatial interpolation technique combined with modified Kriging interpolation and fractal interpolation is used to establish the characteristic spectral interpolation surface related to soil heavy metals and obtain characteristic spectral data by interpolation of coordinate values.And then we apply the inversion model of soil heavy metal content based on spectral features to calculate the heavy metal content data of pending points.In the study area,the recovery error rate of the sampling data is about 30% on average,and the data prediction error rate for unknown samples is about 40%,which indicates that the technical system of the method is practicable and can be used as a chemical analysis method of on-site approximation rapid test and monitoring.In summary,the technology system for rapid monitoring of heavy metal pollution based on hyperspectral technology is feasible in this study.It provides a technical route and guidance for the rapid monitoring of soil heavy metal content,and also provides a reference for other rapid detection based on hyperspectral features.