基于GA-SVM算法的烤烟香型自动识别研究
|
摘要
为了对不同香型烤烟进行特征差异性识别,选取清香型、浓香型和中间香型3类香型的514个烟叶样品,对其中的68种致香成分进行检测,结合数据分析和模式识别技术,提出了一种基于烟草致香成分和遗传算法-支持向量机(GA-SVM)算法的烤烟香型自动识别方法,通过使用遗传算法对支持向量机进行参数优化和调整,并采用5折交叉验证的方法计算分类正确率。分别对GA-SVM算法、SVM算法和朴素贝叶斯算法的分类效果进行对比测试,结果表明:3种模式识别方法对3类香型的分类正确率分别为96.40%、78.58%和84.42%,GA-SVM算法显著优于SVM和朴素贝叶斯等传统分类算法。该方法能够为烤烟香型准确识别、烤烟产地溯源、烟叶香型风格定位提供技术支持。
In order to recognize the characteristic difference of flue-cured tobacco of different flavor types, 514 samples of flue-cured tobacco of fresh, robust and medium flavor types were collected, and 68 aroma components in tobacco samples were determined. By adopting data analysis and pattern recognition technology,an automatic recognition method for tobacco flavor type was proposed on the basis of tobacco aroma components and genetic algorithm-support vector machine(GA-SVM) algorithm. Genetic algorithm was used to optimize and adjust the parameters of support vector machine, and a 5-fold cross-validation method was used to calculate the classification accuracy of the proposed method. The classification results of GA-SVM, SVM and naive Bayesian algorithms were compared, the results showed that the flavor type discrimination accuracies of the three algorithms for the samples were 96.40%, 78.58% and 84.42%, respectively; GA-SVM was significantly more accurate than the other two algorithms. The proposed method provides a technical support for the accurate flavor type discrimination and growing area tracing of flue-cured tobacco.
In order to recognize the characteristic difference of flue-cured tobacco of different flavor types, 514 samples of flue-cured tobacco of fresh, robust and medium flavor types were collected, and 68 aroma components in tobacco samples were determined. By adopting data analysis and pattern recognition technology,an automatic recognition method for tobacco flavor type was proposed on the basis of tobacco aroma components and genetic algorithm-support vector machine(GA-SVM) algorithm. Genetic algorithm was used to optimize and adjust the parameters of support vector machine, and a 5-fold cross-validation method was used to calculate the classification accuracy of the proposed method. The classification results of GA-SVM, SVM and naive Bayesian algorithms were compared, the results showed that the flavor type discrimination accuracies of the three algorithms for the samples were 96.40%, 78.58% and 84.42%, respectively; GA-SVM was significantly more accurate than the other two algorithms. The proposed method provides a technical support for the accurate flavor type discrimination and growing area tracing of flue-cured tobacco.
引文
[1]乔学义,申玉军,马宇平,等.不同香型烤烟烟叶香韵研究[J].烟草科技,2014(2):5-7,14.QIAO Xueyi,SHEN Yujun,MA Yuping,et al.Study on characteristic aroma notes of flue-cured tobacco leaves of different flavor styles[J].Tobacco Science&Technology,2014(2):5-7,14.
[2]丁瑞康,王承瀚,朱尊权,等.卷烟工艺学[M].北京:食品工业出版社,1958:49-50,60-71.DING Ruikang,WANG Chenghan,ZHU Zunquan,et al.Cigarette technology[M].Beijing:Food Industry Press,1958:49-50,60-71.
[3]李章海,王能如,王东胜,等.不同生态尺度烟区烤烟香型风格的初步研究[J].中国烟草科学,2009,30(5):67-70,76.LI Zhanghai,WANG Nengru,WANG Dongsheng,et al.Preliminary study of aroma type styles of flue-cured tobacco in different ecological scale regions[J].Chinese Tobacco Science,2009,30(5):67-70,76.
[4]宗浩,杨程,陈刚,等.不同烤烟品种香型风格与多酚类物质含量差异分析[J].中国农学通报,2011,27(30):241-245.ZONG Hao,YANG Cheng,CHEN Gang,et al.The analysis of difference of aroma type and polyphenol content in different flue-cured tobacco varieties[J].Chinese Agricultural Science Bulletin,2011,27(30):241-245.
[5]席元肖,魏春阳,宋纪真,等.不同香型烤烟化学成分含量的差异[J].烟草科技,2011(5):29-33,65.XI Yuanxiao,WEI Chunyang,SONG Jizhen,et al.Content variance of some chemical components in flue-cured tobacco of different flavor styles[J].Tobacco Science&Technology,2011(5):29-33,65.
[6]席元肖,宋纪真,李锋,等.不同香型烤烟香气前体物及香味成分含量的差异分析[J].浙江农业科学,2011(2):355-361.XI Yuanxiao,SONG Jizhen,LI Feng,et al.Analysis on contents of flavor precursors and flavor components of flue-cured tobacco of different aroma types[J].Journal of Zhejiang Agricultural Sciences,2011(2):355-361.
[7]刘金霞,李元实,黄飞,等.不同香型烤烟化学成分含量的差异研究[J].河南农业科学,2012,41(9):50-52.LIU Jinxia,LI Yuanshi,HUANG Fei,et al.Research on difference of chemical components in flue-cured tobacco of different flavor styles[J].Journal of Henan Agricultural Sciences,2012,41(9):50-52.
[8]郭灵燕,袁红星,海洋,等.河南省不同香型烟叶香气成分比较分析[J].河南农业科学,2010(6):40-44.GUO Lingyan,YUAN Hongxing,HAI Yang,et al.Comparison analysis of volatile aroma components in leaves of different flavor types of Henan tobacco[J].Journal of Henan Agricultural Sciences,2010(6):40-44.
[9]王建波,周清明,邓小华,等.湖南浓香型烤烟产区主要化学成分比较[J].湖南农业科学,2013(9):31-34,38.WANG Jianbo,ZHOU Qingming,DENG Xiaohua,et al.Comparison of main chemical components between full flavor flue-cured tobacco leaves from different main tobacco production areas of Hunan Province[J].Hunan Agricultural Sciences,2013(9):31-34,38.
[10]李伟,陈江华,詹军,等.烤烟香型间致香物质组成比例及其差异分析[J].中国烟草学报,2013,19(2):1-6.LI Wei,CHEN Jianghua,ZHAN Jun,et al.Differential analysis of aroma component proportion in various aroma style flue-cured tobacco leaves[J].Acta Tabacaria Sinica,2013,19(2):1-6.
[11]杜咏梅,张建平,王树声,等.主导烤烟香型风格及感官质量差异的主要化学指标分析[J].中国烟草科学,2010,31(5):7-12.DU Yongmei,ZHANG Jianping,WANG Shusheng,et al.Major chemical indices leading the difference among different flavor types and sensory quality grades of flue-cured tobacco[J].Chinese Tobacco Science,2010,31(5):7-12.
[12]王刘胜,马戎.浓香型产区烟叶主要化学成分与风格品质特色及其关系研究[J].中国烟草科学,2013,34(5):28-32.WANG Liusheng,MA Rong.Major chemical components and quality and style and their relationships in full flavor tobacco growing regions[J].Chinese Tobacco Science,2013,34(5):28-32.
[13]薛超群,王建伟,奚家勤,等.烤烟烟叶理化指标与浓香型风格程度的关系[J].烟草科技,2012(1):52-56.XUE Chaoqun,WANG Jianwei,XI Jiaqin,et al.Relationship between physical-chemical indexes and full flavor style degree of flue-cured tobacco leaves[J].Tobacco Science&Technology,2012(1):52-56.
[14]常爱霞,张建平,杜咏梅,等.烤烟香型相关化学成分主导的不同产区烟叶聚类分析[J].中国烟草学报,2010,16(2):14-19.CHANG Aixia,ZHANG Jianping,DU Yongmei,et al.Cluster analysis of flue-cured tobacco leaves from different growing areas according to the chemical components correlating with aroma types[J].Acta Tabacaria Sinica,2010,16(2):14-19.
[15]詹军,周芳芳,邓国宾,等.基于化学成分和致香物质的烤烟上部叶香型判别分析[J].湖南农业大学学报(自然科学版),2013,39(3):232-241.ZHAN Jun,ZHOU Fangfang,DENG Guobin,et al.Discriminant analysis of aroma types of upper leaf in flue-cured tobacco based on chemical constituents and aroma components[J].Journal of Hunan Agricultural University(Natural Sciences),2013,39(3):232-241.
[16]周泽弘,曹淋海,王昌全,等.基于RBF神经网络建立库存烟叶香型的预测模型[J].中国烟草科学,2016,37(2):65-70.ZHOU Zehong,CAO Linhai,WANG Changquan,et al.The establishment of prediction model of inventory tobacco flavor based on RBF neural network[J].Chinese Tobacco Science,2016,37(2):65-70.
[17]郭东锋,姚忠达,徐迎波,等.基于化学成分的烤烟香型分类研究[J].湖南文理学院学报(自然科学版),2013,25(4):85-89.GUO Dongfeng,YAO Zhongda,XU Yingbo,et al.Study on the classification of flue-cured tobacco based on chemical components[J].Journal of Hunan University of Arts and Science(Natural Science),2013,25(4):85-89.
[18]Vapnik V N.The nature of statistical learning theory[M].Berlin:Springer,1995.
[19]Vapnik V N.An overview of statistical learning theory[J].IEEE Transactions on Neural Networks,1999,10(5):988-999.
[20]Burges C.A tutorial on support vector machines for pattern recognition[J].Data Mining and Knowledge Discovery,1998,2(2):121-167.
[21]Smola A J,Sch?lkopf B.A tutorial on support vector regression[J].Statistics and Computing,2004,14(3):199-222.
[22]李航,赵海东,申金媛,等.基于BPSO和SVM的烟叶近红外有用特征光谱选择[J].物理实验,2015,35(6):8-12.LI Hang,ZHAO Haidong,SHEN Jinyuan,et al.Screening the effective features in the near-infrared spectroscopy of tobacco leaf based on BPSO and SVM[J].Physics Experimentation,2015,35(6):8-12.
[23]宋相中,赖衍清,李祖红,等.近红外光谱结合LSTSVM算法判别不同等级烟叶[J].分析科学学报,2014,30(3):327-331.SONG Xiangzhong,LAI Yanqing,LI Zuhong,et al.Grading tobacco leaves based on near infrared spectroscopy combined with least square twins Support Vector Machine[J].Journal of Analytical Science,2014,30(3):327-331.
[24]Li H M,Wu J D,Huang K K,et al.Study of flue-cured tobacco classification model based on the PSO-SVM[J].Research Journal of Applied Sciences,Engineering and Technology,2013,5(19):4671-4676.
[25]向金海,杨申,樊恒,等.基于稀疏表示的烤烟烟叶品质分级研究[J].农业机械学报,2013,44(11):287-292.XIANG Jinhai,YANG Shen,FAN Heng,et al.Grading for tobacco leaf quality based on sparse representation[J].Transactions of the Chinese Society for Agricultural Machinery,2013,44(11):287-292.
[26]章平泉,杜秀敏,金岚峰,等.支持向量机方法在烟叶可用性预测中的应用[J].中国烟草科学,2011,32(4):19-22.ZHANG Pingquan,DU Xiumin,JIN Lanfeng,et al.Application of Support Vector Machine method in tobacco leaf usability forecast[J].Chinese Tobacco Science,2011,32(4):19-22.
[27]王涛.SVM在配方感官评估中的应用[J].微计算机信息,2010,26(4-1):236-238.WANG Tao.Application on SVM in formulating sensory evaluation[J].Control&Automation,2010,26(4-1):236-238.
[28]魏中华.基于t假设检验及SVM神经网络的卷烟真伪判定[J].烟草科技,2015,48(2):75-78.WEI Zhonghua.Method for cigarette authenticity discrimination based on t test and SVM neural network[J].Tobacco Science&Technology,2015,48(2):75-78.
[29]聂铭,周冀衡,杨荣生,等.基于MIV-SVM的烤烟评吸质量预测模型[J].中国烟草学报,2014,20(6):56-62.NIE Ming,ZHOU Jiheng,YANG Rongsheng,et al.MIV-SVM-based prediction model for smoking quality of flue-cured tobacco[J].Acta Tabacaria Sinica,2014,20(6):56-62.
[30]王玉,王保兴,武怡,等.卷烟挥发性成分的聚类分析[J].烟草科技,2007(2):48-52.WANG Yu,WANG Baoxing,WU Yi,et al.Cluster analysis of volatile components in cigarette[J].Tobacco Science&Technology,2007(2):48-52.
[31]Cevikalp H.New clustering algorithms for the support vector machine based hierarchical classification[J].Pattern Recognition Letters,2010,31(11):1285-1291.
[32]杨志民,刘广利.不确定性支持向量机-算法及应用[M].北京:科学出版社,2012.YANG Zhimin,LIU Guangli.Uncertainty support vector machine:algorithm and application[M].Beijing:Science Press,2012.
[33]米凯利维茨Z.演化程序-遗传算法和数据编码的结合[M].周家驹,何险峰,译.北京:科学出版社,2000.Michalewicz Z.Genetic algorithms+data structuresevolution programs[M].ZHOU Jiaju,HE Xianfeng,translator.Beijing:Science Press,2000.
[2]丁瑞康,王承瀚,朱尊权,等.卷烟工艺学[M].北京:食品工业出版社,1958:49-50,60-71.DING Ruikang,WANG Chenghan,ZHU Zunquan,et al.Cigarette technology[M].Beijing:Food Industry Press,1958:49-50,60-71.
[3]李章海,王能如,王东胜,等.不同生态尺度烟区烤烟香型风格的初步研究[J].中国烟草科学,2009,30(5):67-70,76.LI Zhanghai,WANG Nengru,WANG Dongsheng,et al.Preliminary study of aroma type styles of flue-cured tobacco in different ecological scale regions[J].Chinese Tobacco Science,2009,30(5):67-70,76.
[4]宗浩,杨程,陈刚,等.不同烤烟品种香型风格与多酚类物质含量差异分析[J].中国农学通报,2011,27(30):241-245.ZONG Hao,YANG Cheng,CHEN Gang,et al.The analysis of difference of aroma type and polyphenol content in different flue-cured tobacco varieties[J].Chinese Agricultural Science Bulletin,2011,27(30):241-245.
[5]席元肖,魏春阳,宋纪真,等.不同香型烤烟化学成分含量的差异[J].烟草科技,2011(5):29-33,65.XI Yuanxiao,WEI Chunyang,SONG Jizhen,et al.Content variance of some chemical components in flue-cured tobacco of different flavor styles[J].Tobacco Science&Technology,2011(5):29-33,65.
[6]席元肖,宋纪真,李锋,等.不同香型烤烟香气前体物及香味成分含量的差异分析[J].浙江农业科学,2011(2):355-361.XI Yuanxiao,SONG Jizhen,LI Feng,et al.Analysis on contents of flavor precursors and flavor components of flue-cured tobacco of different aroma types[J].Journal of Zhejiang Agricultural Sciences,2011(2):355-361.
[7]刘金霞,李元实,黄飞,等.不同香型烤烟化学成分含量的差异研究[J].河南农业科学,2012,41(9):50-52.LIU Jinxia,LI Yuanshi,HUANG Fei,et al.Research on difference of chemical components in flue-cured tobacco of different flavor styles[J].Journal of Henan Agricultural Sciences,2012,41(9):50-52.
[8]郭灵燕,袁红星,海洋,等.河南省不同香型烟叶香气成分比较分析[J].河南农业科学,2010(6):40-44.GUO Lingyan,YUAN Hongxing,HAI Yang,et al.Comparison analysis of volatile aroma components in leaves of different flavor types of Henan tobacco[J].Journal of Henan Agricultural Sciences,2010(6):40-44.
[9]王建波,周清明,邓小华,等.湖南浓香型烤烟产区主要化学成分比较[J].湖南农业科学,2013(9):31-34,38.WANG Jianbo,ZHOU Qingming,DENG Xiaohua,et al.Comparison of main chemical components between full flavor flue-cured tobacco leaves from different main tobacco production areas of Hunan Province[J].Hunan Agricultural Sciences,2013(9):31-34,38.
[10]李伟,陈江华,詹军,等.烤烟香型间致香物质组成比例及其差异分析[J].中国烟草学报,2013,19(2):1-6.LI Wei,CHEN Jianghua,ZHAN Jun,et al.Differential analysis of aroma component proportion in various aroma style flue-cured tobacco leaves[J].Acta Tabacaria Sinica,2013,19(2):1-6.
[11]杜咏梅,张建平,王树声,等.主导烤烟香型风格及感官质量差异的主要化学指标分析[J].中国烟草科学,2010,31(5):7-12.DU Yongmei,ZHANG Jianping,WANG Shusheng,et al.Major chemical indices leading the difference among different flavor types and sensory quality grades of flue-cured tobacco[J].Chinese Tobacco Science,2010,31(5):7-12.
[12]王刘胜,马戎.浓香型产区烟叶主要化学成分与风格品质特色及其关系研究[J].中国烟草科学,2013,34(5):28-32.WANG Liusheng,MA Rong.Major chemical components and quality and style and their relationships in full flavor tobacco growing regions[J].Chinese Tobacco Science,2013,34(5):28-32.
[13]薛超群,王建伟,奚家勤,等.烤烟烟叶理化指标与浓香型风格程度的关系[J].烟草科技,2012(1):52-56.XUE Chaoqun,WANG Jianwei,XI Jiaqin,et al.Relationship between physical-chemical indexes and full flavor style degree of flue-cured tobacco leaves[J].Tobacco Science&Technology,2012(1):52-56.
[14]常爱霞,张建平,杜咏梅,等.烤烟香型相关化学成分主导的不同产区烟叶聚类分析[J].中国烟草学报,2010,16(2):14-19.CHANG Aixia,ZHANG Jianping,DU Yongmei,et al.Cluster analysis of flue-cured tobacco leaves from different growing areas according to the chemical components correlating with aroma types[J].Acta Tabacaria Sinica,2010,16(2):14-19.
[15]詹军,周芳芳,邓国宾,等.基于化学成分和致香物质的烤烟上部叶香型判别分析[J].湖南农业大学学报(自然科学版),2013,39(3):232-241.ZHAN Jun,ZHOU Fangfang,DENG Guobin,et al.Discriminant analysis of aroma types of upper leaf in flue-cured tobacco based on chemical constituents and aroma components[J].Journal of Hunan Agricultural University(Natural Sciences),2013,39(3):232-241.
[16]周泽弘,曹淋海,王昌全,等.基于RBF神经网络建立库存烟叶香型的预测模型[J].中国烟草科学,2016,37(2):65-70.ZHOU Zehong,CAO Linhai,WANG Changquan,et al.The establishment of prediction model of inventory tobacco flavor based on RBF neural network[J].Chinese Tobacco Science,2016,37(2):65-70.
[17]郭东锋,姚忠达,徐迎波,等.基于化学成分的烤烟香型分类研究[J].湖南文理学院学报(自然科学版),2013,25(4):85-89.GUO Dongfeng,YAO Zhongda,XU Yingbo,et al.Study on the classification of flue-cured tobacco based on chemical components[J].Journal of Hunan University of Arts and Science(Natural Science),2013,25(4):85-89.
[18]Vapnik V N.The nature of statistical learning theory[M].Berlin:Springer,1995.
[19]Vapnik V N.An overview of statistical learning theory[J].IEEE Transactions on Neural Networks,1999,10(5):988-999.
[20]Burges C.A tutorial on support vector machines for pattern recognition[J].Data Mining and Knowledge Discovery,1998,2(2):121-167.
[21]Smola A J,Sch?lkopf B.A tutorial on support vector regression[J].Statistics and Computing,2004,14(3):199-222.
[22]李航,赵海东,申金媛,等.基于BPSO和SVM的烟叶近红外有用特征光谱选择[J].物理实验,2015,35(6):8-12.LI Hang,ZHAO Haidong,SHEN Jinyuan,et al.Screening the effective features in the near-infrared spectroscopy of tobacco leaf based on BPSO and SVM[J].Physics Experimentation,2015,35(6):8-12.
[23]宋相中,赖衍清,李祖红,等.近红外光谱结合LSTSVM算法判别不同等级烟叶[J].分析科学学报,2014,30(3):327-331.SONG Xiangzhong,LAI Yanqing,LI Zuhong,et al.Grading tobacco leaves based on near infrared spectroscopy combined with least square twins Support Vector Machine[J].Journal of Analytical Science,2014,30(3):327-331.
[24]Li H M,Wu J D,Huang K K,et al.Study of flue-cured tobacco classification model based on the PSO-SVM[J].Research Journal of Applied Sciences,Engineering and Technology,2013,5(19):4671-4676.
[25]向金海,杨申,樊恒,等.基于稀疏表示的烤烟烟叶品质分级研究[J].农业机械学报,2013,44(11):287-292.XIANG Jinhai,YANG Shen,FAN Heng,et al.Grading for tobacco leaf quality based on sparse representation[J].Transactions of the Chinese Society for Agricultural Machinery,2013,44(11):287-292.
[26]章平泉,杜秀敏,金岚峰,等.支持向量机方法在烟叶可用性预测中的应用[J].中国烟草科学,2011,32(4):19-22.ZHANG Pingquan,DU Xiumin,JIN Lanfeng,et al.Application of Support Vector Machine method in tobacco leaf usability forecast[J].Chinese Tobacco Science,2011,32(4):19-22.
[27]王涛.SVM在配方感官评估中的应用[J].微计算机信息,2010,26(4-1):236-238.WANG Tao.Application on SVM in formulating sensory evaluation[J].Control&Automation,2010,26(4-1):236-238.
[28]魏中华.基于t假设检验及SVM神经网络的卷烟真伪判定[J].烟草科技,2015,48(2):75-78.WEI Zhonghua.Method for cigarette authenticity discrimination based on t test and SVM neural network[J].Tobacco Science&Technology,2015,48(2):75-78.
[29]聂铭,周冀衡,杨荣生,等.基于MIV-SVM的烤烟评吸质量预测模型[J].中国烟草学报,2014,20(6):56-62.NIE Ming,ZHOU Jiheng,YANG Rongsheng,et al.MIV-SVM-based prediction model for smoking quality of flue-cured tobacco[J].Acta Tabacaria Sinica,2014,20(6):56-62.
[30]王玉,王保兴,武怡,等.卷烟挥发性成分的聚类分析[J].烟草科技,2007(2):48-52.WANG Yu,WANG Baoxing,WU Yi,et al.Cluster analysis of volatile components in cigarette[J].Tobacco Science&Technology,2007(2):48-52.
[31]Cevikalp H.New clustering algorithms for the support vector machine based hierarchical classification[J].Pattern Recognition Letters,2010,31(11):1285-1291.
[32]杨志民,刘广利.不确定性支持向量机-算法及应用[M].北京:科学出版社,2012.YANG Zhimin,LIU Guangli.Uncertainty support vector machine:algorithm and application[M].Beijing:Science Press,2012.
[33]米凯利维茨Z.演化程序-遗传算法和数据编码的结合[M].周家驹,何险峰,译.北京:科学出版社,2000.Michalewicz Z.Genetic algorithms+data structuresevolution programs[M].ZHOU Jiaju,HE Xianfeng,translator.Beijing:Science Press,2000.