基于中红外光谱技术结合化学计量学对假冒和掺假梅花鹿茸粉的快速鉴别  

Rapid identification of counterfeit and adulterated Sika deer antler powder based on mid-infrared spectroscopy combined with chemometrics

作　　者：李萌 杨承恩 郭瑞雪 陈维佳 李艳春 李雨婷[5] LI Meng;YANG Cheng-En;GUO Rui-Xue;CHEN Wei-Jia;LI Yan-Chun;LI Yu-Ting(Modern Agricultural College of Changchun Polytechnic,Changchun 130504,China;College of Plant Protection,Jilin Agricultural University,Changchun 130118,China;College of Chinese Materia Medica,Jilin Agricultural University,Changchun 130118,China;Comprehensive Service Center of Haotmangha Township,Qianguerros Mongolian Autonomous County,Songyuan 131527,China;College of Life Science,Jilin Agricultural University,Changchun 130118,China)

机构地区：[1]长春职业技术学院现代农学院,长春130504 [2]吉林农业大学植物保护学院,长春130118 [3]吉林农业大学中药材学院,长春130118 [4]吉林省松原市前郭尔罗斯蒙古族自治县浩特芒哈乡综合服务中心,松原131527 [5]吉林农业大学生命科学学院,长春130118

出　　处：《食品安全质量检测学报》2024年第18期94-100,共7页Journal of Food Safety and Quality

基　　金：吉林省重大科技专项(20220304002YY);吉林省农业关键核心技术示范推广(产业技术体系)项目(202300801-04)。

摘　　要：目的基于中红外光谱技术结合化学计量学方法,建立梅花鹿茸粉快速鉴别模型,识别假冒及掺假梅花鹿茸粉。方法采集黑龙江、吉林、辽宁3省梅花鹿茸、马鹿茸、驯鹿茸样品,整只粉碎,得梅花鹿茸粉、马鹿茸粉、驯鹿茸粉样品各60份,并按照质量百分比5%、10%、20%、40%、60%和80%将驯鹿茸粉掺入到梅花鹿茸粉中,获得掺假梅花鹿茸粉样品60份,共计240份样品。样品进行红外光谱扫描,采用K-S法划分样本;利用5种光谱预处理方法结合支持向量机建模,确定最优光谱预处理方法;应用竞争性自适应重加权算法、连续投影算法提取光谱特征信息;建立支持向量机、随机森林、极限学习机识别模型,比较各方法建模效果,确定最优识别模型。结果各样品组红外光谱略有差异,其中纯梅花鹿茸粉组在1029 cm^(-1)附近多糖C-O伸缩振动吸收较强,纯驯鹿茸粉组在3310 cm^(-1)附近氨基酸N-H伸缩振动吸收较强;纯马鹿茸粉组在1659 cm^(-1)附近蛋白质C=O伸缩振动吸收最弱,而各掺假组样品在此波长的红外吸收随着掺假比例增加而增强。多元散射校正-连续投影算法-支持向量机模型对训练集和测试集的识别率均为100%,且仅需11个特征吸收波长点即可完成建模,识别准确率最高,鉴别速度最快,为最佳识别模型。结论中红外光谱结合化学计量学方法可准确、高效、无损的鉴别出假冒、掺假的梅花鹿茸粉。Objective To establish a rapid identification model for genuine and adulterated powder of Sika deer antler using mid-nfrared spectroscopy technology and chemometrics methods to identify counterfeit and adulterated powder of Sika deer antler.Methods Samples of Sika deer antler,horse antler and reindeer antler were collected from Heilongjiang,Jilin and Liaoning Provinces,and crushed whole to obtain 60 samples of Sika deer antler powder,horse antler powder and reindeer antler powder.According to the mass percentage of 5%,10%,20%,40%,60% and 80%,the reindeer antler powder was mixed into the Sika deer antler powder,and 60 samples of the adulterated Sika deer antler powder were obtained,a total of 240 samples.All the samples were scanned by the infrared spectra,and subdivided by the K-S method.The optimal spectral preprocessing method were determined by 5 spectral preprocessing methods combined with support vector machine model.The competitive adaptive reweighting algorithm and continuous projection algorithm were used to select the characteristic spectral bands of all the samples.The recognition model of support vector machine,random forest and extreme learning machine was established,and the modeling effect was compared to determine the optimal recognition model.Results The infrared spectra of each sample group were slightly different,among which the polysaccharide C-O telescopic vibration absorption was stronger in the pure Sika deer antler powder group near 1029 cm^(-1),the amino acid N-H telescopic vibration absorption was stronger in the pure reindeer antler powder group near 3310 cm^(-1);the protein C=O telescopic vibration absorption in the pure horse deer powder group near 1659 cm^(-1) was the weakest,whereas the infrared absorption of samples of each adulteration group at this wavelength increased with the increase of the proportion of adulteration.The recognition rate of the multivariate scattering correction-continuous projection algorithm-support vector machine model was 100% for both the correction and ve

关 键 词：梅花鹿茸 中红外光谱法 化学计量学 假冒 掺假 

分 类 号：O657.33[理学—分析化学] TS218[理学—化学]