以可見光/近紅外光光度計檢驗羊奶攙雜牛奶

Abstract

市售羊乳的價格幾乎是牛乳的兩倍。常有羊乳攙入牛乳或標示成份不清楚的問題，為了維持市場秩序及保護消費者的權益，需要發展一套快速檢測羊奶攙入牛乳的方法。於40。C的溫度下，利用分光光度計(FOSS NIRSystem 6500)掃描随機取樣台大農場生產的羊奶及牛奶，60羊奶樣本各摻雜15% 、10%、6%牛奶，波長從400nm到2500nm，使用光譜間距為2nm，即一次掃描可得1050個吸收光譜值，用WinISIⅡ軟體輸出，再用Unscramber 7.6軟體部分之最小平方迴歸與Matlab 6.0軟體之區別分析法搭配虛擬迴歸技術、建立校正模式。 原始光譜全在波長400nm~2498nm以PLSR配合虛擬迴歸分析，驗證純羊奶樣本對15%摻雜樣本有97.5%準確率，對10%摻雜樣本有72.5%驗證率，對6%摻雜樣本有72.5%驗證率。 以區別分析選取40個相關係數較低波長對應之光譜值，純羊奶與15%、10%及6%的摻雜牛奶樣本，驗證率分別為90%、85%及70%。三組的平均驗證率為81.7%高於PLSR分析的結果80.8%。建立驗證率與羊奶摻雜牛奶比率成線性之方程式可用耒推估未耒量測羊奶摻雜牛奶比例之驗證率。

Market price of goat milk is about twice of milk. To solve the problems of goat milk mixed milk and unclear labeling the compositions ,to maintain market order and to protect consumers' rights and interests, it is necessary to develop a rapid detecting method. At temperature 40 degree C, 60 random goat samples were taken from the experimental farm of National Taiwan University, mixed with 15%,10% and 6% of cow , scanned by FOSS NIRSystem 6500 from 400nm to 2498nm with 2nm intervals.1050 absorbance were collected by winISIⅡ software, and analyzed by PLSR in Unscramber 7.6 software and discriminant analysis in Matlab 6.0 software with a dummy regression technique. Using original spectra from 400nm to 2498nm analyzed by a PLSR method with a dummy regression, validation rates of 60 goat milk samples mixed with 15% , 10%,and 6%cow milk were 97.5%,72.5% and72.5%, respectively. Selecting 40 absorbance corresponding to wavelength with the lowest correlation analyzed by discriminant analysis, validation rates of 60 goat milk samples mixed with 15%, 10%, and 6% cow milk were 90%, 85% and 70%, respectively. The average validation rate of three groups was 81.7% higher than 80.8% by PLSR method. The established linear equation could predict the validation rate without measurement.