Summary: | 碩士 === 國立屏東科技大學 === 生物機電工程系所 === 99 === To guarantee the quality of domestic milk that customers bought, this study used spectral technique detecting cow raw milk and goat raw milk adulterated by reconstituted milk. Percentages of test were 0%, 2%, 5%, 10%, 20%, 30%, 50%, 75%, and 100%.Constituents including fat, solids not fat (SNF), protein, lactose, density, total solids (TS), water were recorded. Application spectrum was visible/near-infrared spectroscopy. There were two methods that we used: multiple linear regression (MLR) and modified partial least squares regression (MPLSR) with differential and smoothing proprocess. Selection samples with random in cow milk for training were 204, and testing samples were 103.In goat milk, training samples were 216, and testing samples were 108.In regression of constitutents, the highest R2 for goat milk is 0.96 which is higher than that of cow milk.In percentage prediction, cow milk and goat milk R2 are 0.97 and 0.99 respectively.The minimal detectable ratio of adulteration is 5%.The results of this study can be used for further application.
|