CAJ | 학술논문

利用傅里叶变换近红外光谱仪采用积分球漫反射方式对60个豆浆样品进行光谱的采集，结合常规分析结果分别建立了3种成分的近红外校正模型。结果表明：豆浆蛋白质、脂肪及可溶性固形物光谱分别经过消除常数偏移量、一阶导数和矢量归一化（SNV）预处理后建模效果最好。蛋白质、脂肪和可溶性固形物含量的校正模型决定系数（R2）分别为：0．9664、0．9500和0．9507，交叉验证均方根差（RMSECV）依次为0．0769、0．0874和0．316；对模型进行外部验证，验证集化学值和模型预测值之间差异不显著，说明模型可以用于豆浆中蛋白质、脂肪和可溶性固形物含量的检测。
이용부리협변환근홍외광보의채용적분구만반사방식대60개두장양품진행광보적채집，결합상규분석결과분별건립료3충성분적근홍외교정모형。결과표명：두장단백질、지방급가용성고형물광보분별경과소제상수편이량、일계도수화시량귀일화（SNV）예처리후건모효과최호。단백질、지방화가용성고형물함량적교정모형결정계수（R2）분별위：0．9664、0．9500화0．9507，교차험증균방근차（RMSECV）의차위0．0769、0．0874화0．316；대모형진행외부험증，험증집화학치화모형예측치지간차이불현저，설명모형가이용우두장중단백질、지방화가용성고형물함량적검측。
With the mode of integrating sphere diffuse,the spectra of 60 soy milk samples were obtained by the Fourier transform near -infrared spectrometer (FT -NIRS)in this research.Combined with the results of chemical analysis,the calibration models of the three components were established separately.The calibration models had a best prediction performance when the spectra of the protein,fat and soluble solids were preprocessed by constant off-set elimination,first derivative and standard normal variate transformation (SNV)respectively.The determination co-efficients (R2 )for the protein,fat and soluble solids content were 0.966 4,0.950 0 and 0.950 7 respectively,and the root mean square errors of cross -validation (RMSECV)were 0.076 9,0.087 4 and 0.316 respectively.Exter-nal validation of the model showed there was no significant difference between chemical values and model predictions, which indicated that the calibration models could be used to detect protein,fat and soluble solids content of soy milk.