Aiming at disadvantages of labor and time consuming of traditional tea classification method, this paper proposes a tea variety classification algorithm that integrates spectral and image characteristics with Mengding Mountain Huangya tea, Zhuyeqing tea and Ganlu tea of Ya'an City, Sichuan Province as objects. It firstly collected the hyperspectral images of tea samples with “GaiaSorter” hyperspectral sorter. After performing relevant pretreatment, 18 spectral characteristic parameters including red-edge position, absorbing area and absorbing depth were extracted according to the spectral curves and 28 image characteristics including average gray scale, consistency and entropy were extracted according to the images. The confluent characteristics were carried out dimensionality reduction with PCA (principal component analysis) method before they are classified and identified with C-SVM algorithm. Experimental results showed that classification of three varieties of tea can be realized rapidly when the input principal component is selected as 3 and the accuracy rate of identification is up to 100%.
[1]
Xu Hai-we.
Application of BP neural network in classification of fresh tea grade based on machine vision
,
2014
.
[2]
Yong He,et al.
[Identification of green tea brand based on hyperspectra imaging technology].
,
2014,
Guang pu xue yu guang pu fen xi = Guang pu.
[3]
Sheng Ye-hua.
Study on feature extraction and experiment of hyperspectral data
,
2008
.
[4]
H. Cai,et al.
[Estimation of Fraction of Absorbed Photosynthetically Active Radiation for Winter Wheat Based on Hyperspectral Characteristic Parameters].
,
2015,
Guang pu xue yu guang pu fen xi = Guang pu.
[5]
赵杰文 Zhao Jiewen,et al.
Estimation of Tea Quality Level Using Hyperspectral Imaging Technology
,
2008
.
[6]
YU Ying-ji,et al.
Identification of Tieguanyin Tea Grades Based on Hyperspectral Technology
,
2014
.