In order to reduce divergences in the interpretation of the results obtained by current methods of fraud detection in powdered coffees, a new method to quantify coffee adulterants is presented in this study. This method is based on image analyses and on the principle that different organic materials found in ground coffees present distinct spectral signatures. The hypothesis is that the adultered coffees, when submitted to an artificial source of illumination, present different reflectance values in RGB (Red/Green/Blue) channels in comparison to those from pure coffees. After the cleaning, drying, and homogenization steps, multispectral images of coffee samples were generated by using a glass magnifier connected to a CCD (Charge Coupled Device) camera. The images were stored and classified by means of a digital image software. The spectral signatures of each component in the coffee sample were identified from their histograms in percentage, and the amounts (by weight) of adulterants in the coffee sample were obtained by calibration curves. The advantages of this method are the capability to analyse a great number of samples in a relatively short time period, the lack of subjectivity in the results, and the non destruction of the coffee samples. This method reaches a minimum accuracy of 95% in the quantification of adulteration in the powdered coffee.
coffee; powders; adulteration; quality controls; spectrometry
