Research On Inversion Of Soil Heavy Metal Cd Content Based On EEMD And Wavelet Decomposition And Reconstruction

With the rapid advancement of industrialization,the soil environment is facing a major threat,and soil heavy metal pollution has become a serious environmental problem.With the development of remote sensing technology,the use of quantitative remote sensing to monitor soil heavy metal pollution has become a new development trend.However,how to extract information about heavy metal elements from soil spectral information and use this as a basis to carry out quantitative inversion research on soil heavy metal content has become very important.This paper takes the cultivated soil in a certain area of Hengyang as the research object,chooses wavelet decomposition and reconstruction and ensemble empirical mode decomposition(EEMD)two time-frequency analysis methods,and uses the correlation coefficient method to obtain the results after wavelet decomposition and reconstruction and EEMD decomposition.The characteristic bands are extracted from the reconstructed signal,and compared with the commonly used first-order differential,second-order differential and other spectral preprocessing methods,the wavelet decomposition and reconstruction and EEMD decomposition methods are explored for their ability to extract weak information in the soil reflectance spectrum.Based on this,a quantitative inversion model of heavy metal Cd element is constructed.The main conclusions of this study are as follows:1.Wavelet decomposition and reconstruction,EEMD decomposition methods can effectively extract the weak information in the spectrum.In exploring whether the wavelet decomposition and reconstruction,EEMD decomposition method can separate the weak information and noise in the high-frequency signal,and then extract the effective information in the spectrum,it is found that the extraction effect of the weak information and the wavelet basis function are very large The relationship is that different wavelet bases have different decomposition capabilities,and choosing a suitable wavelet base function can achieve the extraction of weak information in the spectrum.For the soil reflectance spectrum in this study area,db4 wavelet can realize the extraction of soil reflectance spectrum information;the EEMD decomposition method can also effectively realize the extraction of weak information in the soil reflectance spectrum.2.Using wavelet decomposition and reconstruction,EEMD decomposition methods to directly extract effective information from the original spectrum can effectively improve the correlation between the spectrum and the Cd content,and improve the accuracy of its inversion model.The comparative analysis is based on the accuracy of the PLS model established based on wavelet decomposition and reconstruction,EEMD decomposition method,first-order differential and second-order differential,and the spectral information about Cd element directly extracted based on wavelet decomposition and reconstruction,EEMD decomposition method is constructed.The accuracy of the model is significantly higher than that of the inversion model constructed by the original spectrum;after the first-order differential enhancement processing is performed on the extracted characteristic bands,the accuracy of the constructed PLS inversion model is significantly improved,which is higher than that of the original signal.The accuracy of the inversion model constructed after the first-order differential processing.3.Studies have shown that the PLS regression model can effectively retrieve the heavy metal Cd content of the soil.In this study,the PLS modeling method was used to establish the soil heavy metal Cd content inversion model.The best inversion model is the PLS model constructed by the db4-R-FD and Cd content of the characteristic band extracted after db4 wavelet decomposition and reconstruction.The modular set R2c reached 0.71,the RMSEC was 0.0021,the verification set R2p was 0.47,and the RMSEP was 0.0062.The second is the PLS model constructed by the characteristic bands EEMD-R-FD and Cd content extracted after EEMD decomposition.The modeling set R2c reached 0.70,the RMSEC was 0.0020,the verification set R2p was 0.49,and the RMSEP was 0.0066.In this study,wavelet decomposition and reconstruction,EEMD decomposition and other methods are introduced into the processing of soil heavy metal measured hyperspectral data,which provides a new idea for hyperspectral inversion of soil heavy metal Cd element content,which can be used to achieve a large range of remote sensing satellite hyperspectral data The heavy metal pollution monitoring provides reference and reference.