Study on Identification of Counterfeit Salmon Meat Based on Infrared Spectroscopy
WU Ting1, ZHONG Nan1*, YANG Ling2
1. College of Engineering, South China Agricultural University,Key Laboratory of Key Technology on Agricultural Machine and Equipment(South China Agricultural University), Ministry of Education,Guangdong Provincial Key Laboratory of Food Quality and Safety, Guangzhou 510642,China
2. School of Information Science and Technology,Zhongkai University of Agriculture and Engineering,Guangdong Province Research Center of Food Safety and Intelligent Control Engineering Technology, Guangzhou 510225,China
摘要: 国内三文鱼市场鱼龙混杂,假冒问题严重,但鉴别方法有限。采用红外光谱技术结合偏最小二乘判别分析法(PLS-DA)研究了黑龙江大马哈鱼、淡水虹鳟、智利太平洋鲑三种鱼肉对挪威三文鱼的冒充问题。采用FITR光谱仪和KBr压片法采集四种肉类的原始光谱,并对原始光谱分别进行多元散射校正(MSC)、Savitzky-Golay平滑、一阶导数(first derivative)、标准正则变换(SNV)、峰面积归一化(peak area normalization)五种预处理来消除噪声等干扰因素并确定最佳预处理方法。为建立PLS-DA鉴别模型,将四种鱼肉的光谱分别赋予-3,-1,1和3四个参考分值,建模后通过预测检测集鱼肉得分来检验模型准确性。结果表明:采用峰面积归一化法时,PLS-DA检测模型的效果最好,校正集和交叉验证集的决定系数分别为0.97和0.95。RMSEC和RMSECV分别为0.37和0.52。该模型能显著区分四种鱼肉、检测集的预测分值分别聚集在各自的参考分值附近,在阈值为±1的判别条件下预测准确度为96%。同时采用马氏距离法进一步对四种鱼肉的光谱进行分析,发现相互之间差异明显,其中挪威三文鱼与其品种差别最大的淡水虹鳟距离最大,与其比较接近的智利太平洋鲑的距离最小,红外光谱信息能够反映不同鱼肉的品种、生活环境等差异。因此,采用红外光谱技术结合PLS-DA法能够准确的鉴别出其他鱼肉对挪威三文鱼的冒充问题,同时对其他肉类检测有一定借鉴意义。
关键词:红外光谱;三文鱼;假冒鉴别;偏最小二乘判别分析;马氏距离
Abstract:There are various salmons of different varieties and prices at domestic salmon market. As Norwegian salmon can be sold at higher prices, the counterfeit problem of Norwegian salmon has become extremely serious in China. However, the identification methods are limited and it’s hard for consumers to identify different salmons. Therefore, a quick and accurate identification method is urgently needed. In the paper, infrared spectroscopy and partial least squares discriminant analysis (PLS-DA) were used to detect Norwegian salmon and three counterfeit Norwegian salmons (Heilongjiang salmon, freshwater rainbow trout, Chile Pacific salmon). The study used Fourier transform infrared (FITR) spectrometer and KBr pressed pellet method to acquire the original spectrum of the above four salmons. To eliminate interference factors such as noise and particles scattering, original spectra were preprocessed with multiplicative scatter correction (MSC), Savitzky-Golay, first derivative, standard normal variate (SNV) and peak area normalization respectively and the impact of each method on the model was studied as well. To establish PLS-DA model, spectra of the above four salmons were assigned reference score -3, -1, 1, 3 respectively at threshold range 1 to test the accuracy of the model. The results showed that PLS-DA model achieved the best performance when peak area normalization was used, with determination coefficient of calibration sets 0.97, cross validation sets 0.96, root mean square error of calibration sets(RMSEC) 0.37 and root mean square error of cross validation(RMSECV) 0.52. The model could significantly distinguish four salmons and achieved 96% accuracy when test sets was predicted. Mahalanobis distance method was also used in the study to find the differences of four kinds of spectra. The results showed Norwegian salmon had largest mahalanobis distance with rainbow trout and had smallest mahalanobis distance with Chile Pacific salmon, which was consistent with the difference of salmons’ species and living environments. In summary, infrared spectroscopy combined with PLS-DA is proved to be an effective method for detecting counterfeit salmons and the method may provide a way to distinguish other kinds of meat.
吴 霆,钟 南,杨 灵. 红外光谱技术的三文鱼肉假冒鉴别[J]. 光谱学与光谱分析, 2017, 37(10): 3078-3082.
WU Ting, ZHONG Nan, YANG Ling. Study on Identification of Counterfeit Salmon Meat Based on Infrared Spectroscopy. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2017, 37(10): 3078-3082.
[1] JIANG Jian-jun, DENG Cai, LI Hua(江建军,邓 材,李 华). Food and Machinery(食品与机械), 2012, 26(6): 40.
[2] FEI Xue(飞 雪). China Science & Technology Education(中国科技教育), 2015, 9: 62.
[3] Kuswandi B, Putri F K, Gani A A, et al. Journal of Food Science and Technology, 2015, 52(12): 7655.
[4] Pu H, Xie A, Sun D W. Food and Bioprocess Technology, 2015, 8(1): 1.
[5] Guillén A, Del Moral F G, Herrera L J. Neural Computing and Applications, 2010, 19(3): 465.
[6] De La Roza-Delgado B, Soldado A, De Faria Oliveira A F G, et al. Food Analytical Methods, 2014, 7(1): 151.
[7] LIU Wei-dong, MAO Xiao-ting, JIN Huai-zhou, et al(刘卫东, 毛晓婷, 金怀洲, 等). Journal of China Jiliang University(中国计量学院学报), 2015, 2: 177.
[8] JIANG Yi-li, WU Xiao-zong, HAO Li-hua, et al(蒋祎丽, 吴晓宗, 郝莉花, 等). Food Research and Development(食品研究与开发), 2015, 36(21): 133.
[9] Kamruzzaman M, Makino Y, Oshita S, et al. Food and Bioprocess Technology,2015,8(5): 1054.
[10] TAO Lin, WU Zhong-chen, ZHANG Peng-yan, et al(陶 琳, 武中臣, 张鹏彦, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2011, 27(5): 364.
[11] ZHANG Zhi-qiang, NIU Zhi-you, ZHAO Si-ming(张志强, 牛智有, 赵思明). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2011, 27(11): 388.
[12] WAN Peng, PAN Hai-bing, LONG Chang-jiang, et al(万 鹏, 潘海兵, 龙长江, 等). Food and Machinery(食品与机械), 2012, 6: 164.
[13] WANG Jing-rong, ZHANG Zhuo-yong, ZHANG Zhen-wei, et al(汪景荣, 张卓勇, 张振伟, 等). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2016, 36(2): 316.