Discrimination of white teas produced from fresh leaves with different maturity by near-infrared spectroscopy

White tea is a special tea product with increasing market demand. The assessment of white tea quality is mainly based on panel sensory by sensory evaluation experts, which is time costly and is limited by many uncertainties. This study established a rapid and accurate method for classification of white teas produced from buds and young leaves and that produced from mature leaves and shoots using near-infrared spectroscopy (NIR). Back propagation neural network modelling and support vector machine (SVM) modelling were compared with six pre-processing methods. The best performance was provided by SVM with particle swarm optimization combined with Savitzky-Golay filter pre-processing method, achieving the accuracy of 98.92% in test samples. The NIR-related chemical compounds of two categories of white teas produced from fresh leaves with different maturity were analyzed, including catechins, alkaloids, amino acids and flavonol glycosides. Compared with chemical component concentration, NIR absorbance had a distinct advantage in quick classification of white teas based on the principal components analysis. In addition, the sensory characteristics of two categories white teas produced from fresh leaves with different maturity were also assessed by panelist. The result showed that characteristics of "umami-like" and "smooth" were more likely present in white teas produced from buds and young leaves, while "woody" and "coarse" characteristics were usually present in white teas produced from mature leaves and shoots. Thus, NIR technique is a rapid and reliable method for discrimination of white teas produced from fresh leaves with different maturity, and is a potential method to discriminate sensory characteristics of white teas.