| pH is one of the signifificant properties of soil,it not only affects soil fertility,structure,and microbial diversity,but also affects the form,transformation,and availability of soil nutrients.Therefore,it is very important to study soil pH for the rational use,improvement and management of soil.Soil pH is not only related to the chemical elements of soil,and on the other hand,but also inextricably linked to the organic acids of soil.Therefore,this study uses Laser Induced Breakdown Spectroscopy(LIBS)technology,Mid-infrared spectroscopy(MIR)and LIBS-MIR spectroscopy data fusion technology to quantitatively analyze pH of soil.However,due to the diversity and complexity of the chemical composition of soil samples,the collected LIBS and MIR spectra are very complex.With the help of chemometrics technology,effective information can be extracted to the greatest extent from complex spectral data,thereby improving the accuracy of quantitative analysis results.Based on the actual needs of rapid and accurate determination of soil pH,this study aims to establish a method for measuring soil pH by combining LIBS technology,MIR technology and LIBS-MIR spectral data fusion technology with random forest(RF)algorithm.This method provides a new idea and method for rapid and accurate determination of soil pH.The dissertation is divided into four chapters,the main research contents as follows:Firstly,taking 21 soil samples as research objects,the effects of different input variables(full spectrum,different spectral ranges,the intensity of characteristic bands and characteristic lines)on the predictive performance of RF calibration model were explored.The parameters of calibration model were optimized based on Out-of-bag errors estimation.A method for measuring soil pH based on LIBS spectroscopy technology and RF algorithm was established.The results show that the RF calibration model has a good predictive performance.The internal verification results are:R2=0.9995,RMSE=0.0201;the external verification results are:R2=0.9687,RMSE=0.1285.This method has the advantages of rapidness and accuracy,and provides a new idea and method for the quantitative analysis of soil pH.Secondly,the influence of three different spectral preprocessing methods(First Derivative,Multivariate Scattering Correction,Standard Normal Variation)on the predictive performance of RF calibration model were explored.Input variables were optimized via variable importance thresholds.A soil pH measurement method based on MIR technology and RF algorithm was established.The results show that the model has a good predictive performance.The internal verification results are:R2=0.9887,RMSE=0.0875;the external verification results are:R2=0.9208,RMSE=0.1476.This method has the advantages of rapidness and accuracy,and provides a new idea and method for the quantitative analysis of soil pH.Thirdly,with the help of data fusion strategy,the soil pH measurement methods based on LIBS-MIR spectral data fusion technology combined with RF algorithm was established.First of all,the RF calibration model based on low-level fusion of LIBS-MIR spectral data was constructed,and the predictive results were compared with the RF calibration model constructed based on the optimal LIBS and MIR spectra.Then,the RF calibration model based on the mid-level fusion of LIBS-MIR spectral data was constructed to predict the pH of soil.Finally,the predictive performance of RF calibration models based on different fusion methods were compared.The results show that the RF calibration model based on mid-level spectral data fusion has better predictive performance.The internal verification results are:R2=0.9997,RMSE=0.0163;the external verification results are:R2=0.9809,RMSE=0.0645.Compared with the single spectrum analysis method based on LIBS or MIR,this method has more accurate analysis results,and provides a new ideas and methods for rapid and accurate quantitative analysis of soil pH. |
PDF Full Text Request