Abstract

In this work, a novel and very interesting analytical methodology based on coupling of digital image processing and three-way calibration has been developed for determination of nitrite in food samples. Nitrite in contact with Griess reagent is able to produce a red-colored azo dye whose color intensity is correlated with nitrite concentration and here, a piece of Whatman filter paper impregnated with Griess reagent was used as the platform of the sensor and a SONY Xperia Z5 cell phone was used for image capturing from the sensor surface. To generate second-order data, the F-number of the camera's sensor was changed as an instrumental parameter. Two calibration models were constructed by unfolded partial least squares-residual bilinearization (U-PLS/RBL) and multiway-PLS/RBL (N-PLS/RBL) and then, their performance for prediction of nitrite concentration in test samples was evaluated and the results confirmed a good performance for U-PLS/RBL (REP = 3.25 ppm, RMSEP = 8.82 ppm, RMSEC = 4.62 ppm, Q(2) = 0.99, gamma(-1) = 0.05 and LOD = 0.1 ppm) which was better than that for N-PLS/RBL (REP = 13.98 ppm, RMSEP = 37.86 ppm, RMSEC = 6.46 ppm, Q(2) = 0.98, gamma(-1) = 0.07 and LOD = 0.15 ppm) in predicting concentration of nitrite in test samples which motivated us to choose it for the analysis of cabbage, carrot, lettuce, watermelon, onion, potato, kielbasa and sausage as real samples.