期刊名称：International Journal of Multimedia and Ubiquitous Engineering
印刷版ISSN：1975-0080
出版年度：2016
卷号：11
期号：10
页码：131-142
出版社：SERSC
摘要：In this paper raw milk was taken as the research object, using mid-infrared spectrum analysis method rapidly tests protein and fat content 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 protein and fat, and calibration sets contain 64 samples, and validation sets contain 20 samples. Then, denoising by second derivative method which window size is 17 points, and selecting method by comparing SPA and GA characteristic wavelength respectively, 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 former is higher and characteristic wavelength is least, and its characteristic wavelength point reduces from 220 to 4, R2 of protein model validation set is 0.8010929, RMSEP is 0.0207505, R2 of fat model validation set is 0.91007989, RMSEP is 0.045329, relative standard deviation RSD<3%. These prove that the method for rapid testing of raw milk components is feasible
