, and after it the principal component analysis (PCA) method ( Wold et al., 1987 ; Martens and Næs, 1991 ), the cluster analysis (CA) method ( Heise and Winzen, 2002 ), and the polarqualificationsystem (PQS) method ( Kaffka and Seregély, 2002 ) were
Time consuming and expensive methods have been applied for detection of coffee adulteration based on the literature. In the present work, an optical method (vision system) and the application of an electronic tongue is proposed to reveal the addition of barley in different proportion to coffee in ground and brewed forms. In a range of 1 to 80% (w/w) Robusta coffee was blended with roasted barley. Principal Component Analysis (PCA) accomplished on vision system image data showed a good discrimination of the adulterated samples. The results of Polar Qualification System (PQS) data reduction method revealed even small differences in the right barley content order by point method approach. With Partial Least Squares (PLS) regression the amount of barley in Robusta was predicted with high R2 (0.996) and relatively low RMSEP (∼2%) values in case of vision system data processing. Considering electronic tongue measurements, PCA results showed a good discrimination of samples with higher barley concentration. Misclassification was found in the low concentrated area by Lienar Discriminant Analgsis (LDA). To obtain an accurate model for barley content prediction in coffee, the most sensitive sensor signals were used to apply PLS regression successfully (R2=0.97, RMSEP=3.99% (w/w)).
In the industrial practice Howard mould count (Howard, 1911) is used for the estimation of mould contamination of foods. It was developed originally for the investigation of mould contamination of tomato purée. It is also used nowadays for quality control purposes for other food products as well. Recently this constitutes the basis of the acceptance of the finished products in international trade. This technique demands experts with a lot of practice and morphological proficiency. The investigation makes use of eyes and so the tiredness of the investigator can cause uncertainties.The possibility of other methods for the determination of mould contamination of tomato purée was investigated to replace the Howard method. The NIR technique — as a rapid, non-destructive, reagentless and accurate method — was anticipated as a suitable method for the mentioned purpose. Canned tomato purée had been allowed to become mouldy then the sample was blended with non-mouldy samples in different ratio, so a series of tomato purées containing known amounts of mouldy purée was prepared. Howard mould counts and ergosterol content — another mould contamination relating value — was used as reference for NIR calibration.At quantitative investigation better results were obtained using ergosterol values. The best correlation coefficient (R=0.93) and the smallest standard error of calibration (SEC=0.008 mg g−1 ergosterol) was achieved with triangular smoothing and second derivation of the spectra. At qualitative investigation Polar Qualification System (PQS) was used. Clusters between samples with low and high ergosterol levels could be separated.