Identification of tea color by using computer vision

To overcome the deficiency in tea sensory evaluation such as the result subjectivity and poor coherence, a new method of tea identification was proposed based on soft independent modeling of class analogy pattern recognition theory. By using computer vision to quantitatively depict tea color characters, three predictive models for Biluochun, Longjing, and Qihong teas were built. Under the α=5% significance level, the three predictive models were the best. The results showed that, the back estimation rates for own class samples and rejection rates for other class samples of three models are all 100% in training; the identification rates of three models for own class samples are 90%, 90%, and 100% respectively in prediction, while the rejection rates for other class samples of three models are all 100%. The experimental results showed that the method is feasible with computer vision identifying tea categories based on the tea color characteristics.