Abstract:

Hyperspectral technology combined with support vector machine (SVM) as a classification theory was applied to
identify the grades of Tieguanyin tea. Twenty characteristic spectral parameters were extracted based on the hyperspectral
data of tea samples, including red edge amplitude, blue edge position, yellow edge area, red valley reflectivity, normalized
difference vegetation indexes, etc. The optimal values for the penalty parameter (C) and the kernel parameter (g) were
determined based on the SVM classification model with the radial basis function (RBF) as the kernel function by using
these characteristic spectral parameters as the inputs. An identification model for Tieguanyin tea grades was constructed
and verified. The best experimental results were obtained using the RBF SVM classifier with C = 106 and g = 0.007 5. The
discrimination accuracy rate for unknown Tieguanyin tea samples was 92.86%, suggesting that hyperspectral technology can
be utilized for rapid, nondestructive and accurate identification of Tieguanyin tea grades.

Key words: hyperspectral technology, support vector machine (SVM), Tieguanyin, grade identification