Environmental origin classification of coffee beans using infrared spectroscopy

Abstract

Coffee is one of tropical agricultural products cultivated in many counties and consumed by people worldwide. The main purpose of this study is to employ the infrared spectroscopy technique to rapidly classify the environmental origins of green coffee bean samples. To achieve this purpose, diffuse reflectance spectral data of coffee samples were collected and acquired in wavelength rang of 1000 – 2500 nm. Classification models were established using principal component analysis (PCA) combined with linear discriminant analysis (LDA). The result showed that coffee bean sample can be classified based on their environmental origins with maximum total explained variance of the first two principal components is 97% (PC1 87% and PC2 10% respectively). Judging from the confusion matrix of the LDA, the classification accuracy reach 92%. It may conclude that infrared spectroscopy approach can be used to rapidly classify and sort coffee beans based on their geographical and environmental origins.