ABSTRACT
As a secondary analysis method, reproducibility and reliability of near-infrared spectroscopy (NIRS) quantitative analysis are quite dependent on modelling process. In this paper, it is focused on outlier analysis for protein quantitative model of wheat based on NIRS. The purpose is to discuss the outlier effect in modelling process of complex sample set. The indicator of outliers is the deviation between two interpretative percentage curves in partial least squares regression (PLSR) modelling, when two percentage curves have significant deviation or departure point, the sample set should include the outliers. The innovative research work is the analysis and treatment of outliers. On the basis of sub-model ergodic calculation method, outliers can be gradually identified and picked-up. The standard deviation of model's prediction residual is used as the reference graduation to distinguish the degree of deviation. According to the degree of deviation from sample population, outliers can also be divided into significant outliers, relative outliers and potential outliers. In this paper, the significant outliers of the sample set are about 7.8%, and the relative outliers are about 15.6%. The outliers will pull normal samples apart from the ideal fitting line and make the dispersity increase. No matter modelling with removed outliers or weighted samples, the purpose is to make the fitting results of quantitative analysis modelling more inclined to majority samples, while reducing or eliminating the impact of outliers.
ABSTRACT
The medium temperature black body (MTBB) is conventional high precision equipment used as spectral radiometric scale in infrared spectral region. However, in near-infrared (NIR) spectral region, there are few papers about spectral radiometric calibration by using MTBB, that is because NIR spectral region is the borderland of its effective spectral region. The main research of this paper is spectral radiometric calibration method by using MTBB in NIR spectral region. Accordingly, this paper is devoted mostly to a discussion of how the calibration precision could be affected by selecting different structural parameters of calibration model. The purpose of this paper is to present the results of research and provide technical reference for improving the traceability in NIR spectral radiometric calibration. In this paper, a NIR fiber coupled spectrometer, whose wavelength range covers from 950 to 1700 nm, has been calibrated by a MTBB with adjustable temperature range from 50 to 1050 °C. Concentrating on calibration process, two key points have been discussed. For one thing, the geometric factors of radiation transfer model of the calibration systems have been compared between traditional structure and fiber direct-coupled structure. Because the fiber direct-coupled model is simple and effective, it has been selected instead of traditional model based on the radiation transfer between two coaxial discs. So, it is an advantaged radiation transfer model for radiometric calibration of fiber coupled spectrometer. For another thing, the relation between calibration accuracy and structural parameters of calibration model has been analyzed intensively. The root cause is scale feature of attribute of calibration data itself, which is the nonlinear structure in scales of spectral data. So, the high precision calibration needs nonlinear calibration model, and the uniform sampling for scale feature is also very important. Selecting sample is an inevitable problem when the nonlinear model is explained by small sample dataset. As the analytic results, there are obviously influences for the calibration precision among different strategies of selecting model's structural parameters. The calibration precision, which is mathematical described by standard deviation of spectral data for calibration, could be from ±0.1% to ±1%.
ABSTRACT
To simplify the model and improve the precision of prediction model, latent projective graph (LPG) was used for variable selection. The original spectrum was processed by continuous wavelet transform (CWT), LPG was obtained by principal component analysis (PCA), and based on the assumption that collinear wavelengths might have the same contribution to the modeling, a few latent spectral variables were selected for establishing prediction model. The root mean square error of prediction (RMSEP) model was 0. 3454, better than other modeling methods. This work proved that variable selection with LPG could simplify the near-infrared spectral model effectively, and improve the precision of prediction model.
