원문정보
초록
영어
This paper takes raw milk as research object, and using mid-infrared spectrum analysis method rapidly tests dry matter in raw milk through establishing quantitative analysis model. First of all, researchers collect 84 kinds of raw milk which is different areas and varieties in Heilongjiang province, and classify 84 samples into calibration sets and validation sets by standard chemical testing and spectral scanning on raw milk, and calibration sets contain 64 samples, and validation sets contain 20 samples. Then, respectively selecting SPA and GA characteristic wavelength after treating by smoothing denoising method which window size is 11 points, and establishing PLS model and model demonstration. Eventually, researchers draw the conclusion that SPA wavelength selection method which value of m_max is 10 and GA wavelength selection method which value of Evaluat is 10d can effectively improve validation set model precision, but precision of the latter is higher and characteristic wavelength point reduces from 352 to 10, The calibrating determination coefficient R2 of model is 0.8092155, root mean square error (RMSEC ) is 0.1206172. Validation set determination coefficient R2 is 0.8620867, the root mean square error (RMSEC) is 0.0950656, and relative standard deviation RSD<3%. These prove that the method for rapid testing of raw milk components is feasible.
목차
1. Introduction
2. Theory
2.1. Kennard-Stone (K-S)
2.2. Successive Projections Algorithm(SPA)
2.3. Genetic Algorithm (GA)
3. Materials and Methods
3.1. Preparation of Test Samples
3.2. Sample Classification
3.3. Spectral Acquisition
4. Results and Discussion
4.1. Spectral Denoising
4.2. Characteristic Wavelength Selection
4.3. Quantitative Analysis Model
5. Conclusions
Acknowledgements
References