基于高光谱成像技术快速检测茶叶茶多酚含量

目的 为了快速、无损的检测茶叶中茶多酚含量，建立一种精确、高效的多元校正模型。方法 首先利用高光谱成像技术采集单纵茶叶的光谱数据，其次通过二维相关光谱(two-dimensional correlation spectroscopy techniques，2D-COS)波段筛选算法提取特征光谱，最后结合极限学习机(extreme learning machine，ELM)建立茶多酚的预测模型，并与全波段模型进行对比。结果 经二维相关光谱算法所提取后的特征波段所建立的模型预测效果优于全波段模型。茶多酚的决定系数(correlation coefficient of cross-validation，R2)从0.89上升到0.94，预测值均方根误差(root mean square error of prediction，RMSEP)也从2.37%下降到2.16%。结论 表明二维相关光谱波段筛选算法有效的提取茶多酚的特征波段，对茶叶茶多酚含量的快速、无损预测具有可行性。

Rapid detection of tea polyphenols in tea based on hyperspectral imaging technology

Objective In order to rapidly and nondestructively detect the tea polyphenols content in tea, an accurate and efficient multivariable calibration model was established. Methods Hyperspectral imager (HSI) was used to collect the spectral data of Dancong Tea, and the characteristic spectrum was extracted by 2D-COS variable selection algorithm and combine with the prediction model of tea polyphenols which compared with the full-band model established by extreme learning machine (ELM). Results The prediction effect of the model based on the characteristic bands extracted by the 2D-COS algorithm is better than that of the full-band model. The coefficient of determination (R2) of Tea polyphenols increased from 0.89 to 0.94, and Root Mean Square Error of Prediction (RMSEP) decreased from 2.37% to 2.16%. Conclusion The two-dimensional correlation spectroscopy techniques (2D-COS) variable selection algorithm could effectively extract the characteristic bands of tea polyphenols, and it was feasible for rapid and nondestructive prediction of tea polyphenols content.