Hyperspectral imaging for non-destructive detection of honey adulteration

Abstract

Honey adulteration causes serious economic losses for the industry, and it is difficult to detect various syrup adulteration. There is no doubt that the research on honey adulteration has scientific significance for maintaining the market order of honey and syrup and protecting the legitimate rights and interests of consumers. Hyperspectral images from pure and adulterated samples were captured using a hyperspectral imaging system (400–1000 nm). In this study, hyperspectral imaging and chemometrics were combined to detect honey adulteration, and the prediction model of honey adulteration detection was established. The pure nectar mixed 0 %, 5 %, 10 %, 15 %, 20 %, 30 % and 40 % of adulterants (fructose syrup and sucrose solution). By samples set partitioning based on joint X-Y distances (SPXY), the pure honey samples and the data fusion adulterated samples were assigned to the calibration set (560) and validation set (280) at the ratio of 2:1. The collected hyperspectral images were analyzed by principal component analysis (PCA) to preliminarily detect honey adulteration. Based on effective wavelengths, the adulterated sample analysis model of pure honey and adulterated honey was established. The results showed that the classification accuracy of LIBSVM model for honey adulteration was 92.5 %, which realized the detection of honey adulteration. Partial least square regression (PLSR) was used to establish adulteration level prediction model. The validation accuracy of this model was 0.84 and the root mean square error (RMSEV) of validation was 5.26 %. Therefore, it is feasible to detect honey adulteration by hyperspectral imaging. This method has the advantages of accuracy, simplicity, and greenness.