Yang Dan, Liu Xin, Liu Hong-Gang, Zhang Ying-Bin, Yin Peng
Tea Research Institute of Chinese Academy of Agricultural Sciences, Tea Quality and Supervision Testing Center, Ministry of Agriculture, Hangzhou 310008, China.
Guang Pu Xue Yu Guang Pu Fen Xi. 2013 Jul;33(7):1786-90.
The effect of different resolutions(2, 4, 6, 8, 16 cm(-1)) on the near infrared spectrogram and nitrogen content model for green tea was studied. Test results showed that instrument resolution could influence the spectra quality. The higher the resolution was, the richer the information would be, but the noise would increase. With lower resolution, spectrogram would be much more smooth, but get seriously distorted, and prediction accuracy would decrease at the same time. The partial least squares model was built after spectral pretreatment. When resolution was 4 cm(-1), the RMSEP value of external validation set was 0.054 6, which was obviously lower than others. The Corr. Coeff. was 0.998 2. Its prediction performance was the best and the prediction accuracy better. STDEV and RSD were 0.020 and 0.334 respectively. Resolution 4 cm(-1) for near infrared spectrometer collecting green tea samples was the optimal resolution. This research can provide a reference for parameters selection when collecting green tea spectra with near infrared spectrometer, improve the stability and prediction performance of the model and promote the application and promotion of the near infrared spectroscopy for tea.
研究了不同分辨率(2、4、6、8、16 cm⁻¹)对绿茶近红外光谱图及氮含量模型的影响。试验结果表明,仪器分辨率会影响光谱质量。分辨率越高,信息越丰富,但噪声会增加。分辨率较低时,光谱图会更平滑,但会严重失真,同时预测准确率会降低。在光谱预处理后建立了偏最小二乘模型。当分辨率为4 cm⁻¹时,外部验证集的RMSEP值为0.054 6,明显低于其他值。相关系数为0.998 2。其预测性能最佳,预测准确率较高。标准差和相对标准偏差分别为0.020和0.334。近红外光谱仪采集绿茶样品时4 cm⁻¹的分辨率为最佳分辨率。该研究可为近红外光谱仪采集绿茶光谱时的参数选择提供参考,提高模型的稳定性和预测性能,促进近红外光谱技术在茶叶中的应用与推广。