<![CDATA[Visual identification of green tea types is often inaccurate due to texture similarities and lighting variations, impacting quality assessment, pricing, and consumer trust. This study analyzes the effect of lighting variation on identifying five green tea types using the Gray Level Co-occurrence Matrix (GLCM) and Backpropagation Neural Network (BPNN). A total of 625 dried green tea images were collected under 1-5 LED light setups. Four GLCM features were extracted: contrast, correlation, energy, and homogeneity. Data were split into 80% training and 20% testing sets, then trained using BPNN with a learning rate of 0.3, momentum of 0.04, and a single hidden layer of 25 neurons. Results showed the highest accuracy of 92% occurred under three-lamp lighting. The combination of GLCM and BPNN effectively automates green tea identification, supporting digital quality standardization and enhancing consumer confidence.]]>
Copyrights © 2026
