Application of Effective Wavelengths for Variety Identification of Instant Milk Teas

LIU Fei,WANG Li,HE Yong
DOI: https://doi.org/10.3785/j.issn.1008-973x.2010.03.036
2010-01-01
Abstract:Based on least squares-support vector machine (LS-SVM),the effective wavelength (EW) in visible/near infrared (Vis/NIR) region was proposed as a new approach for the variety discrimination of instant milk teas. This method uses 225 milk tea samples for the calibration set,while 75 samples for the validation set. After partial least squares (PLS) analysis,the EWs were selected according to the X-loading weights and regression coefficients,and an EW-LS-SVM model was developed for the variety discrimination. The PC-LS-SVM model using principal components (PCs) and the WT-LS-SVM model using wavelet transform (WT) were built for comparison. The recognition ratios of calibrations using EW,PC and WT were all 100%,for the calibration set,while 98.7%,98.7% and 100% for the validation set,respectively. An excellent recognition ratio was achieved by these three models. It is feasible to use effective wavelengths in Vis/NIR region for the variety discrimination of instant milk teas and the EW-LS-SVM model can achieve a satisfying recognition ratio.
What problem does this paper attempt to address?