摘要
普洱茶存储年限检测方式存在操作繁琐、分析过程复杂等问题,为实现对不同存储年限的普洱茶的客观、快速的评价,采用电子舌结合离散小波变换-粒子群优化最小二乘支持向量机(DWT-PSO-LSSVM)模型对5种不同存储时间的普洱茶样本进行定性分析。针对电子舌输出信号特点,采用离散小波变换(DWT)作为特征提取方法对输出信号预处理。在此基础上,采用粒子群优化最小二乘支持向量机(PSO-LSSVM)对不同存储年限的普洱茶进行分类鉴别。实验表明,与传统机器学习模型相比,DWT-PSO-LSSVM模型对普洱茶存储年限的分类效果更优,其精确率(Precision)、召回率(Recall)和F1-Score分别达到94.8%、94%和0.936。结果证实,DWT-PSO-LSSVM结合电子舌适合于对普洱茶存储年限进行快速检测,且具有较高的分类准确性。
The detection method of storage year of Pu'er tea has the problems of complicated operation and analysis process.In order to realize fast and objective evaluation of pu'er tea with different storage year,electronic tongue combined with DWT-PSO-LSSVM detection model are adopted to conduct qualitative analysis of tea samples with 5 different storage years.According to the characteristics of the output signal of electronic tongue,discrete wavelet transform is used as the feature extraction method for preprocessing.On this basis,particle swarm optimization-least squares support vector machine(PSO-LSSVM)is used to classify and identify the storage year of Pu'er tea.The experiment shows that compared with traditional machine learning models,DWT-PSO-LSSVM model possesses a better classification effect for pu'er tea,in which the Precision,Recall and F1-Score are 4.8%,94% and 0.936,respectively.The results show that DT-PSO-LSSVM combined with electronic tongue are suitable for fast detection of storage year of Pu'er tea and has high classification accuracy.
作者
荆晓语
缪楠
杨正伟
李庆盛
张鑫
王志强
JING Xiaoyu;MIAO Nan;YANG Zhengwei;LI Qingsheng;ZHANG Xin;WANG Zhiqiang(School of Computer Science and Technology,Shandong University of Technology,Zibo Shangdong 255049,China)
出处
《智能计算机与应用》
2020年第9期86-89,94,共5页
Intelligent Computer and Applications
基金
山东省自然科学基金(ZR2019MF024)
赛尔网络下一代互联网技术创新项目(NGII20170314)
教育部科技发展中心产学研创新基金(2018A02010)。
关键词
普洱茶
电子舌
离散小波变换
粒子群优化算法
最小二乘支持向量机
快速检测
Pu'er tea
electronic tongue
discrete wavelet transform
particle swarm optimization algorithm
least square support vector machines
rapid detection
