Potential of hyperspectral imaging for rapid identification of true and false honeysuckle tea leaves

Abstract

Honeysuckle (Lonicerae japonicae Flos, LJF) tea is a favorite cool tea in China and Southeast Asia. However, some unscrupulous traders usually use Lonicera Flos (LF) as LJF to sell for earning high profit. In order to identify true and false honeysuckle tea leaves rapidly and precisely, hyperspectral imaging technology was applied to develop a nondestructive identification model for LJF and LF. Firstly, the original spectral data were analyzed by three pretreatment methods including Savitzky–Golay (SG) convolution smoothing, multiple scatter correct and standard normal variate transformation (SNV). Then, a full-band analysis model was established by using the partial least squares-discriminant analysis method. And after the selection of characteristic wavelengths by regression coefficients algorithm, the identification analysis models based on the back-propagation neural network and extreme learning machine (ELM) discriminant were established. The results showed that the BP neural network and ELM discriminant analysis model based on SNV denoising at 9 characteristic wavelengths could achieve the best identification results. The recognition rates of both modeling sets and forecasting sets could reach 100%. Therefore, the application of hyperspectral imaging technology can identify LJF and LF effectively and nondestructively, and has potential in the identification of true and false honeysuckle tea leaves.