Main Article Content
Caffeine is an important component in coffee. As well as its stimulatory effects on the consumer, caffeine has many negative effects, which has led to the popularity of caffeine-free coffee throughout the world. Decaffeinating coffee is a laborious and expensive process and only about 40% to 60% successful, but it is often not possible to tell whether or not the label on a jar or packet of coffee has been successfully decaffeinated or not. Therefore, it is possible to fraudulently mislabel decaffeinated coffee. The objective of this study was therefore to assess the caffeine content in blended roasted ground coffee that was made from ground roasted soybeans adding to ground roasted Robusta coffee. Blended roasted ground coffee samples were prepared by adding ground roasted soybean to ground roasted Robusta coffee over the range from 1% to 99% (w/w). All the blended samples were then scanned using a reflectance near infrared hyperspectral imaging (NIR-HSI) spectrophotometer in the wavelength range of 935–1720 nm. Samples (N = 202) were divided into a calibration set (N = 142) and a prediction set (N = 60). A regression model was established for determining caffeine content using partial least squares regression (PLSR). The PLSR model achieved coefficients of determination (R2p) in prediction samples of 0.88 and root mean square error in prediction (RMSEP) of 1.36 mg/g, which indicates a good predictive accuracy for detecting caffeine content in blended roasted ground coffee using NIR-HSI technique. Predictive images were created by interpreting every pixel of spectral image from caffeine content to colors using the model. The visualization of this research revealed that caffeine content could be rapidly predicted related to the color scale. This non-destructive method would be useful to manufacturers for producing new products of blended roasted ground coffee since it would provide a monitoring system thus allowing the level of caffeine to be controlled.