食品与发酵工业 >

2025 , Vol. 51 >Issue 24: 75 - 85

DOI: https://doi.org/10.13995/j.cnki.11-1802/ts.041311

研究报告

机器算法结合光谱主成分特征融合对青稞酒的判别研究

  • 赵玉霞 ,
  • 王茹 ,
  • 张世芝 ,
  • 殷博 ,
  • 张明锦
展开
  • 1(青海师范大学 化学化工学院,青海 西宁,810016)
    2(青海民族大学 化学化工学院,青海 西宁,810016)
    3(青海省环境功能材料先进技术与应用重点实验室,青海 西宁,810016)
第一作者:硕士研究生(张明锦教授为通信作者,E-mail:zhangmingjin@qhnu.edu.cn)

收稿日期: 2024-10-15

  修回日期: 2025-04-30

  网络出版日期: 2026-01-12

基金资助

国家自然科学基金项目(22363010);青海省自然科学基金项目(2022-ZJ-769)

收起

Machine algorithm combined with spectral principal component feature fusion for discriminative study of Qingke liquor

  • ZHAO Yuxia ,
  • WANG Ru ,
  • ZHANG Shizhi ,
  • YIN Bo ,
  • ZHANG Mingjin
Expand
  • 1(College of Chemistry and Chemical Engineering, Qinghai Normal University, Xining 810016, China)
    2(College of Chemistry and Chemical Engineering, Qinghai University for Nationalities, Xining 810016, China)
    3(Qinghai Key Laboratory of Advanced Technology and Application of Environmental Functional Materials, Xining 810016, China)

Received date: 2024-10-15

  Revised date: 2025-04-30

  Online published: 2026-01-12

Fold

摘要

建立基于光谱融合的定性分析模型,实现保护地理标志产品“互助”青稞酒的快速鉴别。采集白酒的紫外光谱(ultraviolet,UV)和近红外光谱(near-infrared,NIR),分别使用4种方法进行预处理,通过主成分特征提取,应用数据层和特征层策略融合多光谱信息,通过比较验证偏最小二乘判别分析(partial least square-discriminant analysis,PLS-DA)、随机森林(random forest,RF)、反向传播神经网络(back propagation neural network,BPNN)和径向基神经网络(radial basis function neural network,RBF-NN)模型的评价指标来评估建模效果。结果表明,二阶导数预处理后主成分特征提取融合的变量建立PLS-DA模型效果最好,预测集的灵敏度、特异性和受试者工作特征(receiver operating characteristic,ROC)曲线下面积(area under the curve,AUC)分别为1.000、0.966 7和0.962 4;原始光谱和Savitzky-Golay平滑(Savitzky-Golay smooth,SG)光谱经过主成分特征提取融合后的变量建立的RF模型最优,训练集和预测集的分类准确率均达到100%;UV原始光谱和SG预处理后经过主成分特征提取的变量建立的BPNN模型识别效果最好,预测集分类准确率和预测决定系数分别为100%和1,均方误差＜0.03;UV原始光谱和SG预处理后的主成分分析-径向基神经网络(principle component analysis-radial basis function neural network,PCA-RBF-NN)分类结果最优,训练集和预测集分类准确率均为100%;NIR全光谱经SNV预处理后建立的RBF-NN模型分类结果最优,训练集和测试集的分类准确率值均为100%;UV-NIR的LF原光谱和SG预处理光谱分类结果最优,训练集和测试集分类准确率均为100%。因此,经主成分特征提取建模所用的光谱数据变量大大减少,有效简化了分类模型,而模型性能仍与全波长所建立的模型性能持平。该文为“互助”青稞酒的快速、无损识别提供了一种可行的方法。

关键词： 中国“互助”青稞酒; 主成分特征提取; 偏最小二乘判别; 反向传播神经网络; 随机森林; 径向基神经网络

本文引用格式

赵玉霞 , 王茹 , 张世芝 , 殷博 , 张明锦 . 机器算法结合光谱主成分特征融合对青稞酒的判别研究[J]. 食品与发酵工业, 2025 , 51(24) : 75 -85 . DOI: 10.13995/j.cnki.11-1802/ts.041311

Abstract

This study established a qualitative analysis model based on spectral fusion to achieve rapid identification of the protected geographical indication product “Huzhu” Qingke liquor. Ultraviolet (UV) and near-infrared (NIR) spectra of Baijiu were collected and preprocessed using four methods. Principal component feature extraction was employed to integrate multispectral information through data layer and feature layer strategies. The modeling effectiveness was evaluated by comparing the performance metrics of partial least square-discriminant analysis (PLS-DA), random forest (RF), back propagation neural network (BPNN), and radial basis function neural network (RBF-NN) models. Results indicated that the PLS-DA model built with variables derived from second derivative preprocessing and principal component feature extraction performed the best, achieving sensitivity, specificity, and area under the curve (AUC) of receiver operating characteristic (ROC) of 1.000, 0.966 7, and 0.962 4 in the prediction set, respectively. The RF model optimized by principal component feature extraction of raw spectra and Savitzky-Golay smooth (SG) spectra achieved the highest classification accuracy of 100% in both training and prediction sets. The BPNN model established with principal component variables from raw UV spectra and SG-preprocessed spectra demonstrated the best recognition performance, with a prediction set classification accuracy of 100% and a prediction coefficient of determination of 1, while the mean squared error (MSE) was less than 0.03. Principal component analysis-radial basis function neural network (PCA-RBF-NN) classification yielded optimal results, achieving 100% classification accuracy in both training and prediction sets. The RBF-NN model built from full NIR spectra after SNV preprocessing also produced the best classification results, with 100% accuracy in both training and test sets. The UV-NIR LF raw spectra and SG-preprocessed spectra classification results were the most optimal, achieving 100% classification accuracy in both training and test sets. Consequently, the spectral data variables used for principal component feature extraction modeling were significantly reduced, effectively simplifying the classification model while maintaining performance parity with models built using full wavelengths. This study provides a feasible method for rapid, non-destructive identification of “Huzhu” Qingke liquor.

Key words： Chinese Huzhu Qingke liquor; principal component feature extraction; partial least square-discriminant analysis; back propagation neural network; random forest; radial basis function neural network

参考文献

[1] 刘建学, 杨国迪, 韩四海, 等.白酒基酒中典型醇的近红外预测模型构建[J].食品科学, 2018, 39(2):281-286.
LIU J X, YANG G D, HAN S H, et al.Prediction model for typical alcohols in base liquor based on near infrared spectroscopy[J].Food Science, 2018, 39(2):281-286.
[2] 孙宝国, 李贺贺, 胡萧梅, 等.健康白酒的发展趋势[J].中国食品学报, 2016, 16(8):1-6.
SUN B G, LI H H,HU X M,et al.The development trend of healthy Baijiu[J].Journal of Chinese Institute of Food Science and Technology, 2016, 16(8):1-6.
[3] 李娜, 程伟, 张杰, 等.白酒原产地分析鉴别技术研究进展[J].酿酒科技, 2018(6):116-121.
LI N, CHENG W, ZHANG J, et al.Research progress in analysis & identification technology of the origin of Baijiu[J].Liquor-Making Science & Technology, 2018(6):116-121.
[4] 凌晨, 马清蓉, 耿超, 等.近红外光谱技术在清香型酒醅检测中的应用研究[J].酿酒, 2024, 51(2):100-106.
LING C, MA Q R, GENG C, et al.Application of near infrared spectroscopy in the detection of Fen flavor fermented grains[J].Liquor Making, 2024, 51(2):100-106.
[5] 苏媛媛, 姜雪, 仓义鹏, 等.紫外-可见光谱传感对高温大曲白酒真实性的准确鉴别[J].化学试剂, 2023, 45(10):8-13.
SU Y Y, JIANG X, CANG Y P, et al. Accurate identification of high-temperature Daqu liquor by UV-vis sensor. Chemical Reagents, 2023, 45(10):8-13.
[6] ARZBERGER U, LACHENMEIER D W.Fourier transform infrared spectroscopy with multivariate analysis as a novel method for characterizing alcoholic strength, density, and total dry extract in spirits and liqueurs[J].Food Analytical Methods, 2008, 1(1):18-22.
[7] 彭秉顺, 李占海.青稞酒工艺特点及产品风格的探讨[J].酿酒科技, 1991(4):25-27.
PENG B S, LI Z H.Discussion on technological characteristics and product style of highland barley wine[J].Liquor-making Science & Technology, 1991(4):25-27.
[8] 许锦文, 李善文.互助青稞酒的香型及其风味特征[J].酿酒科技, 2012(7):82-84;86.
XU J W, LI S W.Investigation on the flavor type and the flavoring characteristics of Huzhu highland barley wine[J].Liquor-Making Science &Technology, 2012(7):82-84;86.
[9] QIAN Y L, AN Y Q, CHEN S, et al.Characterization of Qingke liquor aroma from Tibet[J].Journal of Agricultural and Food Chemistry, 2019, 67(50):13870-13881.
[10] 马华丽, 刘志明, 宋永朋.青海青稞酒挥发性成分的可见-紫外光谱学特性分析[J].中国酿造, 2015, 34(2):158-162.
MA H L, LIU Z M, SONG Y P.Characteristic of UV-Vis absorption spectroscopy about volatile compounds of Qinghai barley liquor[J].China Brewing, 2015, 34(2):158-162.
[11] WANG X L, SONG X B, ZHU L, et al.Unraveling the acetals as ageing markers of Chinese Highland Qingke Baijiu using comprehensive two-dimensional gas chromatography-time of flight mass spectrometry combined with metabolomics approach[J].Food Quality and Safety, 2021, 5:510.1093: fqsafe.
[12] 张世芝, 唐玮琦, 张明锦, 等.基于紫外光谱法的青稞酒快速鉴别方法[J].食品与发酵工业, 2020, 46(14):211-215.
ZHANG S Z, TANG W Q, ZHANG M J, et al.Rapid identification of Qingke liquor based on UV spectroscopy[J].Food and Fermentation Industries, 2020, 46(14):211-215.
[13] CAI J J.PGEToolbox:A Matlab toolbox for population genetics and evolution[J].Journal of Heredity, 2008, 99(4):438-440.
[14] RINGNÉR M.What is principal component analysis?[J].Nature Biotechnology, 2008, 26(3):303-304.
[15] LI C N, QI Y F, SHAO Y H, et al.Robust two-dimensional capped l 2,1-norm linear discriminant analysis with regularization and its applications on image recognition[J].Engineering Applications of Artificial Intelligence, 2021, 104:104367.
[16] WU R Y, HAO N.Quadratic discriminant analysis by projection[J].Journal of Multivariate Analysis, 2022, 190:104987.
[17] ZHANG Z Y, JIANG J Y, WANG G X, et al.Application of two-dimensional correlation UV-vis spectroscopy in Chinese liquor Moutai discrimination[J].American Journal of Analytical Chemistry, 2015, 6(5):395-401.
[18] 胡耀强, 郭敏, 叶秀深, 等.近红外光谱法间接测定白酒酒精度[J].光谱学与光谱分析, 2022, 42(2):410-414.
HU Y Q, GUO M, YE X S, et al.Indirect determination of liquor alcohol content based on near-infrared spectrophotometry[J].Spectroscopy and Spectral Analysis, 2022, 42(2):410-414.
[19] 迟茜, 王转卫, 杨婷婷, 等.基于近红外高光谱成像的猕猴桃早期隐性损伤识别[J].农业机械学报, 2015, 46(3):235-241;234.
CHI Q, WANG Z W, YANG T T, et al.Recognition of early hidden bruises on kiwifruits based on near-infrared hyperspectral imaging technology[J].Transactions of the Chinese Society for Agricultural Machinery, 2015, 46(3):235-241;234.
[20] PENG J T, ZHOU Y C, PHILIP CHEN C L.Region-kernel-based support vector machines for hyperspectral image classification[J].IEEE Transactions on Geoscience and Remote Sensing, 2015, 53(9):4810-4824.
[21] ZHU Q M, CAI Y, LIU L Z.A global learning algorithm for a RBF network[J].Neural Networks, 1999, 12(3):527-540.
[22] 崔永,申加伟,刘洋,等.油压调控系统在挖机瞬变工况下的适应性研究[J].内燃机工程, 2023, 44(5):50-56;65.
CUI Y, SHEN J W, LIU Y, et al.Study on adaptability of oil pressure control system in transient working condition of the excavator[J].Chinese Internal Combustion Engine Engineering, 2023, 44(5):50-56;65.
[23] 余鹏飞,朱继忠,熊小伏,等.基于储能的电力系统安全调控方法[J].电力系统保护与控制,2023,51(19):173-186.
YU P F, ZHU J Z, XIONG X F, et al.Power system safety regulation method based on energy storage[J].Power System Protection and Control, 2023, 51(19):173-186.
[24] 郭刚,汪海涛,高晓成,等.基于粗糙径向基神经网络的刮板输送机负载预测方法研究[J].煤炭工程,2024,56(2):138-145.
GUO G, WANG H T, GAO X C, et al.Research on load forecasting method of scraper conveyor based on rough radial basis function neural network[J].Coal Engineering, 2024, 56(2):138-145.
[25] 李慧,顾洪涛,苏婷婷.近红外光谱技术用于快速检测藜麦营养成分的研究进展[J].农产品加工,2024(1):93-97;102.
LI H, GU H T, SU T T.Research progress of near-infrared spectroscopy for rapid detection of quinoa nutritional components[J].Farm Products Processing, 2024(1):93-97;102.
[26] 许情,吕敏,邓虹霄,等.机器学习在合成大麻素识别鉴定中的应用进展[J].中国药科大学学报, 2024, 55(3):316-325.
XU Q, LYU M, DENG H X, et al.Advances in the application of machine learning in the identification and authentication of synthetic cannabinoids Journal of China Pharmaceutical University, 2024, 55(3):316-325.
[27] 李佳,韩宝瑜,梅献山.基于电子鼻技术的有机绿茶贮存期评价方法探讨[J].茶叶通讯,2024, 51(1):68-77.
LI J, HAN B Y, MEI X S.Study on evaluation method of storage time of organic green tea based on electronic nose technology[J].Journal of Tea Communication, 2024, 51(1):68-77.
[28] DE ALMEIDA V E, DE SOUSA FERNANDES D D, DINIZ P H G D, et al.Scores selection via Fisher's discriminant power in PCA-LDA to improve the classification of food data[J].Food Chemistry, 2021, 363:130296.
Options
文章导航

/

〈 〉

技术支持：北京玛格泰克科技发展有限公司