- 學位論文
應用近紅外線光譜量化國產鮮牛(羊)乳摻入還原乳之研究
Study of Near-infrared Spectroscopy on Quantifying Adulteration Ratio of Reconstituted Fresh Cow Milk and Fresh Goat Milk
摘要
鮮乳是台灣人民主要乳品來源，也是重要的營養來源。為了維護消費者的權利，有必要發展快速檢測鮮乳是否摻入還原乳之技術。本研究透過可見光及近紅外線光譜與乳成分檢測鮮牛乳、未二次加溫鮮羊乳、二次加溫鮮羊乳是否摻入還原乳及其成分之研究，並利用部份最小平方迴歸(PLSR)、多重線性迴歸(MLR)、最小平方支持向量機(LS-SVM) 進行不同摻入比例的光譜與乳成分迴歸分析，來探討不同的迴歸模式分析及隨機取樣方法之準確度。摻入比例為0%、10%、20%、50%。結果顯示，在不同摻入比例迴歸，以LS-SVM模式的最佳數學處理為最佳，在驗證組方面，鮮牛乳、未二次加溫鮮羊乳、二次加溫鮮羊乳之最佳數學處理分別為D16、D12、DD16，其r分別為0.9988、0.9988、0.9989，SEP分別為0.9052、0.9181、0.8697。在不同隨機取樣下摻入迴歸分析方面，顯示都有共同的最佳數學處理。在成分迴歸分析方面，顯示單一成分迴歸比多成分迴歸分析佳。
並列摘要
Fresh milk is a main dairy source for Taiwan people, also an important nutrient source. In order to safeguard the consumers’ rights, there is a need to develop rapid detection technique for fresh milk whether adulterate of reconstituted milk. This study used visible and near-infrared spectroscopy to detect fresh cow milk, fresh goat milk, and reheated fresh goat milk which have adulterated with various reconstituted milk. Partial least squares regression (PLSR), multiple linear regression (MLR) and least squares support vector machine (LS-SVM) were carried on different adulteration ratio spectra and milk component so as to regress the different models and analyze different random sampling on perdiction accuracy. Five tested adulteration ratio were 0%, 10%, 20%, and 50%. The results show, in different adulteration ratio regression models, the LS-SVM model was the best,in the validation group, the fresh milk, fresh goat milk and rewarmed fresh goat milk of best mathematical treatment were D16, D12, and DD16, r were 0.9988, 0.9988, and 0.9989, SEP were 0.9052, 0.9181, and 0.8697. In compositions regression analysis, result suggested that single-compositions perfomed better than multi-compositions did.