高EGCG含量茶树品种光谱识别模型构建

Construction of spectral screening model for tea cultivars with high EGCG content

为解决目前高EGCG含量茶树品种育种过程中,叶片EGCG含量检测存在繁琐、成本高和周期长等问题,研究采用近红外光谱分析技术结合化学计量法实现茶树叶片EGCG含量的快速定性高低判别。利用手持式光谱仪分别采集高EGCG含量(13.68±1.99)%茶树(W1)和低EGCG含量(4.86±1.17)%茶树(Huangdan)叶片的近红外光谱反射率。应用RF(Random Frog)算法提取对EGCG敏感的波段,并对比分析线性判别分析LDA和最小二乘支持向量机LS-SVM两种方法对高EGCG含量茶树叶片的识别能力。结果表明:主成分分析后样本的第一主成分和第二主成分的累计方差贡献率为99.6%,对应的得分对高/低EGCG含量茶树叶片具有较好的聚类作用。RF算法选取前20个敏感波段构建的最小二乘支持向量机(LS-SVM)模型对高/低EGCG含量茶树叶片的识别正确率和Kappa系数分别为93.94%和0.89。研究结果可为高EGCG含量茶树品种的育种提供技术支持。

EGCG content detection was considered a time-consuming,high cost,and cumbersome process during the current high EGCG content tea breeding status.This research was aimed to establish a discriminant model for screening tea cultivar with high EGCG content by combining near-infrared spectroscopy analysis technology and chemometrics.The spectral reflectance of tea leaves from W1 with high EGCG content(13.68±1.99)%and Huangdan with low EGCG content(4.86±1.17)%was collected using a handheld near-infrared spectrometer.The Random Frog(RF)algorithm was applied to select sensitive wavelengths to EGCG.Then two machine learning methods,including linear discriminant analysis(LDA)and least squares support vector machines(LS-SVM),were introduced to gain an optimal model for discriminating tea leaves with high EGCG content.Results from the principal component analysis showed that the first and second principal components(PC1 and PC2)could explain 99.6%information of original data.It presented an excellent clustering effect on high and low EGCG content tea leaves using the scores of PC1 and PC2.The LS-SVM was built based on 20 wavelengths selected by the Random Frog algorithm.It achieved an overall discriminant accuracy of 93.94%and Kappa of 0.89 for tea leaves with high or low EGCG content.The results would provide a guideline for high throughput screening tea cultivar of high EGCG content.