ABSTRACT
In this study, an innovative odor imaging system capable of detecting adulteration in quince seed edible oils mixed with sunflower oil and sesame oil based on their volatile organic compound (VOC) profiles was developed. The system comprises a colorimetric sensor array (CSA), a data acquisition unit, and a machine learning algorithm for identifying adulterants. The CSA was created using a method that involves applying a mixture of six different pH indicators (methyl violet, chlorophenol red, Nile blue, methyl orange, alizarin, cresol red) onto a Thin Layer Chromatography (TLC) silica gel plate. Subsequently, difference maps were generated by subtracting the "initial" image from the "final" image, with the resulting color changes being converted into digital data, which were then further analyzed using Principal Component Analysis (PCA). Following this, a Support Vector Machine was employed to scrutinize quince seed oil that had been adulterated with varying proportions of sunflower oil and sesame oil. The classifier was progressively supplied with an increasing number of principal components (PCs), starting from one and incrementally increasing up to five. Each time, the classifier was optimized to determine the hyperparameters utilizing a random search algorithm. With one to five PCs, the classification error accounted for a range of 37.18% to 1.29%. According to the results, this novel system is simple, cost-effective, and has potential applications in food quality control and consumer protection.
ABSTRACT
A sensitive and simple nanomaterial based colorimetric sensor array (NBCSA) was developed for discrimination of monofloral honey from various European countries based on three botanical origins as Acacia, Canola and Honeydew. The NBCSA was designed by spotting gold (AuNPs) and silver (AgNPs) nanoparticles synthesized using six different reducing and/or capping agents. The colour intensity of AuNPs represented differential changes when interacting with volatile organic compounds appeared in the headspace of the honey samples. The color difference maps, which are calculated as the difference between color intensity of the sensor before and after exposing to the sample vapors, were used as a fingerprint to discriminate the honey samples based on botanical origin. Classification was achieved utilizing data pre-processing and chemometrics data analysis. Fitting accuracies of 88% and 86% were obtained by partial least squares discriminant analysis and linear discriminant analysis whereas 100% was achieved using support vector machine.
