基于GA-SVM算法的烤烟香型自动识别研究

Automatic recognition of flavor types of flue-cured tobacco based on GA-SVM algorithm

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.