A new method to discriminate tea categories

Based on multi-spectral digital image texture feature, a new method for discriminating tea categories was put forward. The images which have three waveband images (Red, Green, NIR) were recorded by multi-spectral digital imager (MS3100). Eight filters were designed based on discrete cosine transform (DCT), and the NIR image was processed by the 8 filters, then the Standard deviation (Sd) of original NIR image and processed NIR image as the input texture feature set for Least squares-support vector machine (LS-SVM) was calculated. One hundred twenty images (twenty for each category) were used for calibration set and one hundred twenty images (twenty for each category) were used as the prediction set in this study. At last, tea categories were classified by LS-SVM. The classification rate using Sd of original NIR image was only 73.33%, while was up to 100% using processed images. The overall results show that the technique combining DCT and LS-SVM can be efficiently utilized for texture recognition of multi-spectral image, and it also is an effective and simple discrimination way for the tea categories.