Detection of adulteration in Chinese honey using NIR and ATR-FTIR spectral data fusion

The aim of this study is to find a fast, accurate, and effective method for the detection of adulteration in honey circulating in the market. Near-infrared spectroscopy and mid-infrared spectroscopy data on natural honey and syrup-adulterated honey were integrated in the experiment. A method for identifying natural honey and syrup-adulterated honey was established by incorporating these data into a Support Vector Machine (SVM). In this experiment, 112 natural pure honey samples of 20 common honey types from 10 provinces in China were collected, and 112 adulterated honey samples with different percentages of syrup (10, 20, 30, 40, 50, and 60%) were prepared using six types of syrup commonly used in industry. The total number of samples was 224. The near- and mid-infrared spectral data were obtained for all samples. The raw spectra were pre-processed by First Derivative (FD) transform, Second Derivative (SD) transform, Multiple Scattering Correction (MSC), and Standard Normal Variate Transformation (SNVT). The above-corrected data underwent low-level and intermediate-level data fusion. In the last step, Grid Search (GS), Genetic Algorithm (GA), and Particle Swarm Optimization (PSO) were employed as the optimization algorithms to find the optimal penalty parameter c and the optimal kernel parameter g for the SVM, and to establish the best SVM-based detection model for natural honey and syrup-adulterated honey. The results reveal that, compared to low-level data fusion, intermediate-level data fusion significantly improves the detection model. With the latter, the accuracy, sensitivity and specificity of the optimal SVM model all reach 100%, which makes it ideal for the identification of natural honey and syrup-adulterated honey.