การตรวจหาความบกพร่องในเมล็ดกาแฟดิบอะราบิกาด้วยเนียร์อินฟราเรด สเปกโทรสโกปีและเคโมเมทริกซ์

Abstract

The purpose of this research was to detect the defects in Arabica green coffee beans by Near Infrared Spectroscopy (NIRS) and chemometrics. Three types of defects were tested, including physically broken, insect damaged and dried cherry. The levels of impurity in green cofiee beans were 0,3, 5, 7, 10, 15, 20, 25, 30, 35, 40, 45 and 50 % by weight, totaling 13 impurity levels per defect type. The spectra were measured using a reflectance mode by the NIRSystem 6500, MPA and Portable NIR, in the range of wavelength of 400-2500, 800-2500 and 900-1700 nm, respectively. The spectral data were exploratorily analyzed using principal component analysis (PCA) to investigate their variations. Partial least squares regression (PLSR) was thereafter employed to develop calibration equations for estimating the impurity levels in green coffee beans. The results revealed that all three NIR spectrometers could detect the defects in Arabica green cofiee beans, the differences of spectral data of each impurity level could be shown by PCA technique. Moreover, the calibration equations could be used to accurately quantify the impurity levels in green cofiee beans, especially NIRSystem 6500 and MPA which are high efficiency. Their predictive results from calibration equations of three defect types have R' and Q ranging from 0.991 1-0.9992 and 0.9395 -0. 9920, respectively, which were considered acceptable for quality assurance. Therefore, the NIRS and chemometrics could be accurately used for detecting the defect in Arabica green coffee beans.