Digital Discrimination On Withering Degree Of Keemun Black Tea Depend On Hyperspectral Imaging

In order to quantitative discrimination on withering degrees,different degrees of withering samples were collected to acquire hyperspectral imaging,and the withering degrees were identified based on near infrared hyperspectral imaging.The contents of catechins and amino acids of samples were detected by high-performance liquid chromatography.The spectral and textural features were extracted from hyperspectral data to discriminate the withering degrees of black tea.Data fusions combined with ratio of catechins to amino acids were used to predict the degrees of black tea.And the hyperspectral imaging technique combined with moisture was utilized to predict the degrees of Keemun black tea.The results showed as the following:(1)Combination of spectrum and image information for discriminating withering degrees of Keemun black tea.The hyperspectral images of all samples were analyzed by principal component analysis(PCA),the first two principal component images were chose according to the variances contribution rate.And then,the five dominant wavelengths(1040,1182,1249,1449 and 1655 nm)were selected as spectral features by the loadings of the first two principal component images under the all wavelengths.Discriminant models were built based on the textural features from first two principal component images and five dominant wavelengths of images,respectively.The results showed that the textural features from five dominant wavelengths of images are more suitable for discriminating withering degrees of Keemun black tea.The SVM model based on data fusion gave the best results with high correct discrimination rate of 94.64%.The results implied that data fusion combined with SVM has the capability of discriminating withering degrees of Keemun black tea.(2)Hyperspectral imaging technique combined with ratio of catechins to amino acids for predicting the degrees of Keemun black tea.Data fusion was utilized to build partial least square(PLS)to predict the ratio of catechins to amino acids.The PLS model showed good performance with the correlation coefficient and root mean square error of the calibration set were 0.8790 and 0.428 %/% respectively,and the correlation coefficient and root mean square error of the prediction set were 0.8805 and 0.43 %/% respectively.The results showed than the hyperspectral imaging technique combined with ratio of catechins to amino acids can be used to predict the degrees of Keemun black tea.(3)Hyperspectral imaging technique combined with moisture for predicting the degrees of Keemun black tea.The SPA-PLS model developed with raw spectra showed the best performance with the correlation coefficient and root mean square error of the calibration set were 0.9651 and 1.87% respectively,and the correlation coefficient and root mean square error of the prediction set were 0.9694 and 1.76% respectively.Subsequently,the SPA-PLS model was utilized to transfer each pixel of the hyperspectral images to its corresponding moisture value for constructing the distribution maps of moisture.