Classification and Recognition of Adulterated Manuka Honey by
Multi-Wavelength Laser-Induced Fluorescence
CHEN Si-ying1, JIA Yi-wen1, JIANG Yu-rong1*, CHEN He1, YANG Wen-hui2, LUO Yu-peng1, LI Zhong-shi1, ZHANG Yin-chao1, GUO Pan1
1. School of Optics and Photonics, Beijing Institute of Technology, Beijing 100081, China
2. Academy of Military Medical Sciences, Academy of Military Sciences, Beijing 100071, China
Abstract：Manuka honey is produced in New Zealand and has strong antibacterial and antioxidant effects. The price is relatively high, and adulteration incidents have occurred frequently in recent years. This paper uses laser-induced fluorescence (LIF) technology to classify and identify Manuka honey adulterated with syrup. Four commonly used lasers of 266, 355, 405 and 450 nm are selected as excitation sources, and three brands of New Zealand Manuka honey (No. A, B and C) adulterated with baking syrup are used as experimental samples. The adulteration ratio ranged from 0% to 90%, with an interval of 10%. Each sample solution has been tested 60 times under different excitation wavelengths, with a total of 7 200 sets of data. For the spectral data, firstly, pretreat with fluorescence band interception, smoothing and normalization; Then, randomly select 80% of the data as training and 20% as test sets. Principal component analysis (PCA) combined with linear discriminant analysis (LDA) is used to reduce the dimension of the training set data; Finally, K-nearest neighbor (KNN) and support vector machine (SVM) classification models are established for the dimensionality reduction data respectively, and the test set data are classified and identified by the models. After 50 times of random grouping and classification calculation, the recognition rate’s average value and standard deviation are obtained. The experimental results show that the excitation wavelength greatly influences the final recognition results. The recognition rate of 266 nm excitation is the highest. The recognition rates of the three Manuka adulterated solutions are more than 98.5%, and the highest can reach 100%; 355 and 405 nm excitation are the second, and the recognition rates of all samples are greater than 92%; The classification effect of 450 nm excitation is the worst, with the recognition rates less than 66%. Therefore, the comparison of classification algorithms only uses the spectral data excited by 266, 355 and 405 nm. The analysis results show that the classification effect of the KNN algorithm is better than the SVM algorithm. For the three honey adulterated solutions excited by 266 nm, the recognition rates of the KNN algorithm are more than 1% higher than that of the SVM algorithm. According to the experimental results, using LIF to classify and identify adulterated Manuka honey is feasible. For Manuka honey adulterated with syrup, among all combinations used in this paper, 266nm excitation combined with PCA-LDA and KNN algorithms has the highest recognition rate and the best classification effect, which provides an effective method for rapid and accurate identification of adulterated Manuka honey.
[1] Hegazi N M, Elghani G, Farag M A. Journal of Food Science and Technology, 2022, 59: 2527.
[2] KANG Hai-ning, CHEN Bo, LAN Fang, et al（康海宁，陈 波，蓝 芳，等）. Modern Food Technology（现代食品科技）, 2019, 35(12): 328.
[3] GE Xue-feng, WU Yan-wei, ZHAN Zhi-min（葛学峰, 吴彦玮, 赵志敏）. Analysis Laboratory（分析试验室）, 2017, 36(6): 660.
[4] Karoui R, Dufour E, Bosset J O, et al. Food Chemistry, 2007, 101(1): 314.
[5] Chen H, Xu Q, Jia Y, et al. Applied Optics, 2021, 60(21): 6140.
[6] ZHAO Jie-wen, HAN Xiao-yan, CHEN Quan-sheng, et al（赵杰文，韩小燕，陈全胜，等）. Spectroscopy and Spectral Analysis（光谱学与光谱分析）, 2013, 33(6): 188.
[7] Bong J, Prijic G, Braggins T J, et al. Food Chemistry, 2017, 214: 102.
[8] Bong J, Loomes K M, Lin B, et al. Food Chemistry, 2017, 267(30): 355.
[9] Szymańska E, Gerretzen J, Engel J, et al. TrAC Trends in Analytical Chemistry, 2015, 69: 34.
[10] ZHOU Zhi-hua（周志华）. Machine Learning（机器学习）. Beijing: Tsinghua University Press（北京：清华大学出版社）, 2016.
