基于遗传算法的安溪铁观音品质快速评价研究  

Rapid Quality Evaluation of Anxi Tieguanyin Tea Based on Genetic Algorithm

作　　者：王冰玉[1] 孙威江[2,3] 黄艳[2] 余文权[4] 吴全金[1] 林馥茗[1] 夏金梅[1] 

机构地区：[1]福建农林大学园艺学院,福建福州350002 [2]福建农林大学安溪茶学院,福建福州350002 [3]福建省茶产业技术开发基地,福建福州350002 [4]福建省农业科学院,福建福州350003

出　　处：《光谱学与光谱分析》2017年第4期1100-1104,共5页Spectroscopy and Spectral Analysis

基　　金：国家质量监督检验检疫总局公益性行业科研专项项目(201410225-4);高等学校博士学科点专项基金项目(20133515110006);国家国际科技合作项目(2010DFB33030-4)资助

摘　　要：为探究一种快速无损的安溪铁观音品质评价方法,利用遗传算法(GA)对茶样的近红外光谱特征波长进行筛选,结合偏最小二乘(PLS),建立全谱段的PLS定量模型与GA-PLS模型。结果表明,傅里叶变换近红外(FT-NIR)全谱段光谱在经过平滑+二阶导数+归一化处理后,PLS模型预测性能最高,建模结果为:校正集相关系数R_C=0.921,校正集均方根误差RMSEC=0.543,验证集相关系数R_P=0.913,验证集均方根误差RMSEP=0.665。选用近红外光谱6 670~4 000cm-1谱区,采用遗传算法进行特征波长筛选,参与建模数据点数从1 557缩减到408个。优选波段后,GA-PLS建模结果为:校正集相关系数R_C=0.959,校正集均方根误差RMSEC=0.413,验证集相关系数R_P=0.940,验证集均方根误差RMSEP=0.587。可见,GA-PLS模型的校正集和验证集的预测结果均优于全谱段PLS模型。结果说明,在传统的近红外光谱技术结合化学计量学方法的建模基础上,加入遗传算法进行波长筛选,能有效提高模型预测能力,实现方法学的创新研究,且GA-PLS品质评价模型具有较强的参考和推广价值,为提高我国茶叶品质的检测技术水平提供新的方法借鉴。Anxi Tieguanyin tea was collected as the research materials in this study. In order to find a fast and non-destructive method for rapid quality evaluation of Anxi Tieguanyin tea, the Genetic Algorithm （GA） was applied to wavelength selection befoe it is combined with partial least squares （PLS） to construct PLS and GA-PLS calibration model. The results showed that the PLS model displayed the highest prediction performance after the Fourier transform near-infrared （FI＇-NIR） spectrum being pro- cessed by smoothing, the second derivative and normalized methods. Statistic results with PLS.. Rc = 0. 921, RMSEC= 0. 543, RP=0. 913, RMSEP=0. 665. NIR spectra ranging from 6 670 to 4 000 cm-1 were selected, and 1 557 data volume for building calibration model were reduced to 408 with Genetic algorithm. Statistic results with GA-PLS： Rc =0. 959, RMSEC=0. 413, Rp =0. 940, RMSEP=0. 587. It has shown that the prediction precision of calibration set and validation set of GA-PLS model is better than those of PLS model. According to the results, it can effectively improve the prediction ability of the model when the Genetic Algorithm （GA） is applied to select the wavelengths in a traditional model which is based on the near infrared spectroscopy combined with partial least squares. It can also achieve the innovation of the methodology. Furthermore, the quality evalua- tion GA-PLS model provides strong reference and possesses promotional value. In addition, it provides valuable reference and new avenue for improving the standard of detection technology of tea quality in China.

关 键 词：近红外光谱 遗传算法 偏最小二乘 安溪铁观音 品质评价 

分 类 号：S571.1[农业科学—茶叶生产加工]