Model Optimization And Stability Of NIR Spectroscopy Analysis Of Organic Matter In Soils

Using the method of near infrared spectroscopy analysis spectroscopy and chemometrics to build and optimize the organic matter of soil mathematical models of quantitative analysis and study the stability of the models. The method is based on the best validation of all the samples of single-point model, which can best divide calibration sample sets and validation sample sets. On this basis, we adopt the single-point regression, PLS, Savitzky-Golay smoothing PLS, the equal interval moving window MLR methods to establish the mathematical models of quantitative analysis, respectively. In the optimizing goal of the root-mean-square error (RMSEP)and the correlation coefficient (R_p), we carry out the optimal design of a variety of relevant model parameters, after a large number of computer numerical experiments carried out in large-scale optimization models and discuss the stability of the models.The results show that: the method of dividing the calibration sample set and validation sample set has a good representation, objectivity and stability; Savitzky-Golay smoothing optimization mode can be greatly improved the predict results of models. the modle of using fixed interval combination discrete wavelength points have best prediction accuracy; The RMSEP, RRMSEP and R_P value of optimal models' prediction effect reaches 0.1451,9.52%,0.9668, respectively, which will provide excellent, stable mathematical models of quantitative analysis to the organic matter of soil with near infrared spectroscopy technology.