| Summary: | 碩士 === 國立屏東科技大學 === 機械工程系 === 94 === The visible/near-infrared spectroscopy and fluorescent image were used in this study to detect the adulteration ratio in fresh and raw milk. Various ratio of adulterated milk (powder milk and fresh milk, powder milk and raw milk) from 0% to 100% were calibrated with three regression mode: MLR, PLSR, MPLSR and Canonical variant analysis (CVA). Results showed that prediction accuracy over than 92% for all regression models. In CVA test, the accuracy in calibration and test were 89.58% and 84.03%, respectively. In the analysis of fluorescent image, 18 features and gray level were tested. The prediction accuracy was 72.97% and 74.50% for 18 features and grey level, respectively.
|
|---|
