부산대학교 도서관

In the world, black tea can be classified as one of the most commercial and popular consumer drinkings. Due to the demand of the world, the tea is adulterated using various additives and colours. As a result, the identification techniques are very important to maintain the quality tea product among the consumers. In China, India, Sri Lanka and many South Asian countries, black tea is a key agricultural export product which brings the main income with millions of dollars each year. Adding sugar is one of the serious problems emerging frequently in the tea industry which severely affect the reputation of the world-famous brands with competition in the world tea market. Therefore researchers have introduced different methods but most of the methods are laborious, expensive and accuracy is not to the expected level. A multispectral imaging system is used in this new technology to detect percentage levels of adulteration. Nine spectral bands with maximum wavelengths between 405 nm to 950 nm were used to design the hardware system. Classification algorithms were developed to identify the adulterant levels initially with machine learning techniques using k-NN (k-Nearest Neighbours), SVM (Support Vector Machines), and deep learning techniques such as FFNN (Feed Forward Neural Network) and CNN (Convolutional Neural Network).