• Spectroscopy and Spectral Analysis
  • Vol. 42, Issue 11, 3336 (2022)
Yi-yun TANG1、*, Rui LIU2、2;, Lu WANG2、2;, Hui-ying LÜ1、1; 4;, Zhong-hai TANG1、1; 4; *;, Hang XIAO1、1; 3;, Shi-yin GUO1、1; 4;, and Wei FAN1、1; 4; *;
Author Affiliations
  • 11. College of Food Science and Technology, Hunan Agricultural University, Changsha 410128, China
  • 22. Baoshan Tobacco Company of Yunnan Province, Baoshan 678000, China
  • show less
    DOI: 10.3964/j.issn.1000-0593(2022)11-3336-09 Cite this Article
    Yi-yun TANG, Rui LIU, Lu WANG, Hui-ying LÜ, Zhong-hai TANG, Hang XIAO, Shi-yin GUO, Wei FAN. Application of One-Class Classification Combined With Spectral Analysis in Food Authenticity Identification[J]. Spectroscopy and Spectral Analysis, 2022, 42(11): 3336 Copy Citation Text show less
    Schematic diagram of different classification methods
    Fig. 1. Schematic diagram of different classification methods
    Simulated data graph ofactual sample
    Fig. 2. Simulated data graph ofactual sample
    Simulated data graph of adulterated samples (three classes)
    Fig. 3. Simulated data graph of adulterated samples (three classes)
    Simulated data graph of adulterated samples (one class)
    Fig. 4. Simulated data graph of adulterated samples (one class)
    真/假PLS-DAOCPLS
    敏感性特异性敏感性特异性
    5/1 0000100%100%0
    50/1 0000100%100%100%
    500/1 000100%83.4%73.8%100%
    1 000/1 000100%75.6%83%100%
    1 000/500100%70.4%83%100%
    1 000/50100%60%83%100%
    1 000/5100%80%83%100%
    Table 1. Classification results of simulated adulterants of different classes
    真/假PLS-DAOCPLS
    敏感性特异性敏感性特异性
    5/1 0000100%100%0
    50/1 0000100%100%100%
    500/1 000100%84.4%73.8%100%
    1 000/1 000100%76.5%83%100%
    1 000/500100%75%83%100%
    1 000/50100%70%83%100%
    1 000/5100%80%83%100%
    Table 2. Classification results of simulated adulterants of one class
    类别检测对象应用技术分析方法掺假物质检测结果参考文献
    食用油奇亚籽油和芝麻油傅里叶变换红外光谱OCPLS,
    SIMCA
    玉米油、 花生油、 大豆油和葵花籽油正确识别率都有94%以上[34]
    茶油近红外光谱和荧光光谱OCPLS菜籽油、 葵花籽油、 玉米油和花生油灵敏度为95.4%, 特异性为91%[35]
    初榨椰子油傅里叶变换衰减全反射红外光谱DD-SIMCA菜籽油、 玉米油、 葵花籽油和大豆油88%~100%的灵敏度, 96%~100%的特异性[36]
    菜籽油傅里叶变换红外光谱SIMCA, PLS-DA,
    DD-SIMCA和
    OCPLS
    玉米油、 花生油、 大豆油和葵花籽油SIMCA, PLS-DA, 的分类效果高于DD-SIMCA和OCPLS[37]
    亚麻籽油近红外光谱OCPLS菜籽油、 玉米油、 葵花籽油, 棉籽油和大豆油正确识别率95.8%[38]
    乳制品牛奶中红外光谱SIMCA甲醛、 过氧化氢、 碳酸氢盐、 碳酸酯和蔗糖82%的正确分类、 17%的不确定分类和1%分类错误[39]
    脱脂奶粉紫外-可见、 荧光和近红外光谱SIMCA和
    OCSVM
    氯化铵、 硝酸铵、 三聚氰胺和尿素总体准确率为86%[40]
    饮品正宗板蓝根茶傅里叶变换近红外光谱OCPLS干苹果皮准确率为93.6%[41]
    猕猴桃汁荧光光谱OCPLS糖浆和人造果粉灵敏度为92.9%[42]
    葡萄蜜酒低场核磁共振光谱OCPLS,
    DD-SIMCA
    和PLS-DA
    苹果汁、 腰果汁和混合果汁分辨率高于93%[43]
    保健品中草药天麻近红外光谱OCPLS芋头淀粉、 甘薯淀粉、 马铃薯淀粉和黄精粉灵敏度为91.07%[44]
    牛至药材近红外光谱PLS-DA,
    DD-SIMCA
    榛子、 橄榄叶和迷迭香等单类分类器功能强大[45]
    燕窝傅里叶变换红外光谱LDA, SVM
    和OCPLS
    银耳、 琼脂、 炸猪皮和蛋清灵敏度为93.7%, 特异度为88.6%[46]
    香辛料辣椒粉傅里叶变换中红外光谱DD-SIMCA苏丹Ⅰ、 苏丹Ⅳ、 铬酸铅、 氧化铅、 二氧化硅、 聚氯乙烯和阿拉伯胶所有掺假物的特异性>80%[47]
    辣椒粉核磁共振光谱DD-SIMCA偶氮红、 甜菜根和漆树粉灵敏度为92%[48]
    谷物木薯淀粉拉曼光谱OCSVM和
    SIMCA
    小麦粉和碳酸氢钠可检测掺假率超过2%的可能性[49]
    杏仁粉高光谱短波红外图像DD-SIMCA花生粉100%的敏感性和89%~100%的特异性[50]
    大豆粕近红外光谱DD-SIMCA三聚氰胺、 氰尿酸和混合掺假物灵敏度为98%[51]
    Table 3. Application of one-class classification combined with spectral analysis in food adulteration detection
    Yi-yun TANG, Rui LIU, Lu WANG, Hui-ying LÜ, Zhong-hai TANG, Hang XIAO, Shi-yin GUO, Wei FAN. Application of One-Class Classification Combined With Spectral Analysis in Food Authenticity Identification[J]. Spectroscopy and Spectral Analysis, 2022, 42(11): 3336
    Download Citation