Research On Inversion Of Heavy Metal Mercury And Chromium Contents In Cultivated Land Soil By Optimized SVM Model

Cultivated soil,as a carrier and natural nutrients for crop growth,provides a basic guarantee for food production,but the emergence of heavy metal pollution seriously harms the quality of the soil environment.In order to quickly obtain and monitor the content of heavy metals in cultivated soils and effectively reflect the status of soil heavy metal pollution,the use of quantitative remote sensing inversion technology for the inversion of heavy metal content in cultivated soils is of great significance for the evaluation of soil environmental quality.This paper selects cultivated soil in a certain area of Hengyang,Hunan Province as the research object,and explores the feasibility of different feature extraction methods and algorithm optimization models for quantitative inversion of the soil heavy metals mercury(Hg)and chromium(Cr)content in the study area.After preprocessing the original spectrum such as smoothing,resampling and spectral transformation,the correlation analysis of soil heavy metal Hg and Cr content is carried out,and the optimal spectral transformation method corresponding to different heavy metal elements is selected,and the iterative retained information variable method(IRIV)is used respectively.Competitive adaptive weighting algorithm(CARS)and Random Frog algorithm(Random Frog)to extract spectral characteristic bands,and finally use support vector machine(SVM),genetic algorithm optimization support vector machine(GA-SVM)and gray wolf algorithm optimization The support vector machine(GWO-SVM)model established the hyperspectral inversion models of the soil heavy metals Hg and Cr content in the study area,and carried out the accuracy evaluation and analysis.The main research results of the thesis are as follows:(1)After spectral preprocessing,the spectral characteristics are obviously prominent,which effectively improves the correlation between the spectral data and the content of heavy metals.Through the overall trend analysis of the correlation curve,the optimal spectral transformation method for heavy metals Hg and Cr elements They are first-order differential and multivariate scattering correction.(2)The spectral characteristic bands of heavy metals Hg and Cr selected under different feature extraction methods can better reduce the redundancy of spectral data,retain effective variable information,reduce model complexity,and improve model prediction accuracy.(3)Through genetic algorithm and gray wolf algorithm optimization,the overall accuracy of support vector machine model inversion has been improved.Among them,the prediction accuracy of heavy metal Hg element content based on IRIV algorithm combined with GA-SVM model is the best,and based on Random Frog algorithm combined The GWO-SVM model predicts the content of heavy metal Cr with the best accuracy.This study uses different feature extraction methods combined with model optimization to quantitatively retrieve the heavy metal content of cultivated land,providing a new idea for the retrieval of heavy metal content of cultivated land in other similar areas,and can provide support for the application of hyperspectral remote sensing images to monitor the heavy metal content of soil on a large scale.