원문정보
초록
영어
The moisture content of rice is of great significance for the eating quality and food safety. Therefore, it is very necessary to establish a rapid, stable, reliable and high prediction accuracy quantitative analysis model that can be used in on-line detection. In this study, the quantitative analysis technique of near infrared diffuse reflectance spectroscopy was used to detect the moisture content in rice. We combine the algorithm (MW-IPLS) based on MWPLS (Moving Window Partial Least Squares) with IPLS (Interval Partial Least Squares) to optimize the characteristic wavelength. Then we establish partial least squares regression in the preferred characteristic wavelength range. The experimental results show that the model of quantitative analysis using the MW-IPLS algorithm to optimize the characteristic wavelength is optimal comparing with the whole spectrum and single method such as MWPLS and IPLS. The numbers of Factors, R2P, RMSECV and RMSEP are 6, 0.8597, 0.2523 and 0.2753 respectively. Therefore, using the WM-IPLS algorithm to optimize the characteristic wavelength can reduce the processing capacity of the data and make the model more concise. In addition, it also provides a new method for the analysis of near infrared spectral characteristic wavelength selection.
목차
1. Introduction
2. Materials and Methods
2.1. Samples Collection, Preparation and Calibration
2.2. Spectrum Acquisition
2.3. Standard Chemical Calibration
3. Results and Discussion
3.1. Spectra Denoising
3.2. Moving Window Partial Least Squares Band Selection
3.3. Interval Partial Least Squares Band Selection
3.4. MW-IPLS Band Selection Model Demonstration
4. Conclusions
References