关键词: 关键词：茶叶, 拼配, 高光谱, 数据融合, 量化

Abstract: Abstract：Blending is one of the important and key steps in export tea processing. In the present situation blending technique completely depends on artificial qualitative analysis (Sensory evaluation). To overcome these issues, we developed the alternative method to evaluate quality of blend tea. Hyperspectral imaging is an emerging technique that integrates conventional imaging and spectroscopy to acquire both spatial and spectral information from a sample. In order to develop the digitized discrimination on quality control of tea blending, different grades of Keemun black tea were collected in our research. In this study we collected around 110 tea blends from Keemun black tea grades 5 and 6 were procured from Xiangyuan Tea Co. Ltd., Qimen country. The relative proportions in tea blends were nondestructively evaluated by hyperspectral imaging technology at the range of 908-1735nm, and we used the hyperspectral imaging system to obtain the spectra and the image information of the tea blends. The characteristic spectra were extracted from the region of interest (ROI), and standard normal variate (SNV) method was preprocessed to reduce background noise. Spectral features were selected by successive projections algorithm (SPA), and all of the hyperspectral images of tea blends samples were analyzed by principal component analysis (PCA). According to the weight coefficient five dominant wavelengths were selected. Textual features were collected by Gray level co-occurrence matrix (GLCM) from five dominant wavelengths of images. Subsequently, partial least squares (PLS), least squares-support vector machine (LS-SVM) and back propagation-artificial neural networks (BP-ANN) classification models were developed based on spectral features, textural features and data fusion, respectively. Compared with the results of the models built with spectral features or textural features, the LS-SVM model based on data fusion showed higher correct discrimination rate in prediction set. The correct discrimination of LS-SVM based on data fusion was 94.5%. The results indicated that hyperspectral imaging combined with LS-SVM was a potent tool in the discrimination of relative proportions in tea blends. The above stated findings were concluded that the hyperspectral imaging technique is fast and scientific of digitized discrimination for relative proportions in tea blends. It provided research basis for the digitization and production standardization of tea blends.

Key words: Key words: tea, blends, hyperspectral imaging, data fusion, quantification