Author(s)

Libo Rao, Xiao Yang, Suzhen Liu, Xiaoyan Chen

Corresponding Author

Xiaoyan Chen

ABSTRACT

This study used hyperspectral imaging technology to achieve rapid identification of varieties of Zanthoxylum bungeanum. Nine optimal wavelengths are extracted from the spectral range of 400-1000 nm by the random frog (RF) algorithm. The texture features of the first, second and third principal component images are extracted with local binary pattern (LBP). K-nearest neighbor (KNN) and support vector machine (SVM) models were established with spectral features and the combination of spectra and textures, respectively. The accuracy of KNN and SVM models based on the combination of spectral and texture are 100%, but the computational efficiency of KNN model is significantly higher than that of SVM model. It provides a fast and accurate theoretical model for the identification of varieties of Zanthoxylum bungeanum.

KEYWORDS

Zanthoxylum Bungeanum, Random Frog, Local Binary Pattern, Texture