The Identification of Green Tea Based on Feature Extraction

Abstract

In order to identify the green tea quickly and accurately, the infrared spectra of teas measured by Fourier transform infrared spectroscopy are analyzed by using informatics method of feature extraction in this work. By comparing the different characteristic bases found by different feature genes of different tea samples, it is concluded that when green tea, yellow tea and white tea are chosen to constitute the standardized spectrum data matrix, and its first and third feature gene are selected to build characteristic basis, the green tea can be visually distinguished and the same samples have a very good aggregation. While when the characteristic basis is built by the first and second feature gene or the second and third feature gene, or the samples which are used to make up of standardized spectrum data matrix are all green tea, the recognition effect of green tea is bad. A linear arbiter of green tea is set up according to neutrality line principle on the above effectively characteristic basis, when all samples are projected onto the characteristic basis it is shown that all the green tea fall on the right of arbiter, and all not green teas are on the left of arbiter, consequently, the green tea can be identified.