國人對於鮮羊乳品質的要求,隨著經濟的發展,愈來愈高。本研究透過可見光與近紅外線光譜,於民國100年12月至102年12月,掃描國產16家品牌鮮羊乳,及於民國102年4月至12月之13條不同路線取樣的生羊乳,探討同一品牌或同一路線在冬季(12-2月)、夏季(7-9月)、及暖季(其餘月份)等三個季節的光譜表現。樣本光譜透過主軸成分分析(Principle Component analysis, PCA)縮小維度,並擷取前三個分數 (Score)數值,利用三維分布圖,以了解同一品牌與同一路線在不同季節樣本間彼此的分佈狀況。另利用多維變異數分析(Multiple Analysis of Variance, MANOVA)及線性判別分析(Linear discriminant analysis, LDA),針對此前三個分數,檢測季節上的差異程度及判別的誤差率。最後應用本實驗室先期所建立的鮮羊乳與生羊乳摻入還原乳檢測模式,包括部份最小平方迴歸(Partial Least Squares Regression, PLSR)模式,與多重線性迴歸(Multiple Linear Regression, MLR)模式,探討國產品牌鮮羊乳與生羊乳在該模式的預測表現。試驗結果顯示,PCA三維分佈圖在鮮羊乳及生羊乳均有明顯分群現象,MANOVA分析鮮羊乳有9個品牌在三種季節中有顯著差異,生羊乳有11條路線在三種季節中有顯著差異。LDA結果得到鮮羊乳最高誤判率為0.72,最低誤判為0.27;生羊乳最高誤判率為0.57最低誤判率為0.2。本研究顯示乳品品質在不同季節是有差異的,此一結果,可作為後續羊乳生產與管理的參考。
Because of economic development, consumers need high quality fresh goat milk more and more. Therefore, it is important to evaluate the quality of goat milk on different seasons, brands, and delivery routes. This study used visible and near-infrared spectroscopy to detect fresh goat milk and raw goat milk. Evaluated factors include three seasons, 16 brands of products, and 13 delivery routes. Sampling dates are from December 2011 to December 2013 on monthly base and sum up to 1570 fresh goat spectra and 585 raw goat milk spectra. Principal Component Analysis (PCA), Multiple Analysis of Variance (MANOVA) and Linear discriminant analysis (LDA) are applied to test their differences. MANOVA test results showed that 9 brands fresh milk and 11 delivery routes raw milk have significant differences in three tested seasons. LDA test results showed that the highest misclassified rate was 0.72 and the lowest was 0.27 for fresh goat milk. For raw goat milk the misclassified rate was 0.57 and 0.2 for the highest and lowest, respectively. These results are useful for dairy goat production and milk quality management.