Remote Sensing Inversion Of Soil Polymetallic Elements Content Based On Stacked Auto-encoder Extreme Learning Machine Model

Remote sensing as a comprehensive technology for earth observation,combines geology,geophysics,geochemistry and other methods,and synthesizes data obtained by different methods to form multiple geoscience data,which is effective in coping with global changes in resources and environment,played a great role in other issues.Based on the theory of material electromagnetic waves,combined with remote sensing data,by analyzing the relationship between the reflectivity of remote sensing data and the content of geochemical elements on the surface,the content and distribution and migration of geochemical elements on the surface are studied,and a high-precision remote sensing geochemical prediction model is constructed.It provides a feasible method for extracting information about geochemical element anomalies.This paper selects the Chenzhou area of Hunan as the research area,and uses the1:200,000 Chenxian breadth geochemical data and Landsat8 OLI remote sensing image as the data,and selects eight kinds of elements in the study area,include Ag,Au,Cu,Mo,Pb,Sn,W,and Zn.As the research object,combined with the characteristics of the enrichment of polymetallic minerals in the study area,the content of polymetallic elements in the study area is inverted.The research content mainly includes the following three aspects:using the R-type factor analysis method to analyze the eight kinds of elements in the study area,select the five factors with the largest variance contribution percentage,representing the five important element combinations in the study area;for geochemical exploration The sampling accuracy of the data and Landsat8 remote sensing images are different.The spatial correspondence between the two data is analyzed to establish a valid data pair;combined with Stacked Auto-encoder(SAE)and Extreme Learning Machine(ELM)algorithms to build four multi-metal element content inversion models,which are a single element inversion model based on ELM,an element combination inversion model based on ELM,and a single element inversion model based on SAE-ELM Based on the SAE-ELM element combination inversion model,this paper uses the average relative error(MAE)and goodness of fit(R~2)to evaluate the accuracy of the model,and predicts the content of polymetallic elements in the study area.The results show that the SAE-ELM element combination inversion model has the highest accuracy,and the distribution of the delineated anomaly target is consistent with the known anomaly region,and the known anomaly region is located in the region or its edge of the inversion prediction data.Finally,area-content multifractal method was used to analyze SAE-ELM element combination inversion results,delineate the lower limit of anomalies,and delineate the high anomaly areas,providing reference for subsequent geological work.