Varietal Classification of Selected Green Coffee Beans (Coffea arabica L. and Coffea canephora Pierre ex A.Froehner) Using Image Processing Software

Abstract

The study created a classification model that sorts green coffee beans based on its variety, primarily ‘Arabica’ and ‘Robusta’ to address coffee adulteration. A total of 1500 green coffee beans were obtained from Cavite, Philippines, allotting 70% and 30% of it for training and testing dataset, respectively. These were then captured through the fabricated image acquisition setup. Image processing techniques were performed to extract the features of the beans namely color, size, shape, and crack using ImageJ software. Fourier analysis was mainly performed for the extraction of the bean shape information. After performing t-test, 3, 4, and 7 parameters of color, size, and shape features were found to be significant, respectively. The 14 features were used in creating 7 classifications setups of different feature combinations through discriminant analysis. The model that yielded the highest accuracy (99%) is the combination of color and size features. This exceeded the subjective varietal classification accuracy (98%) performed by a coffee expert. Thus, the variety of green coffee beans were feasibly classified through image analysis and the created model had surpassed the traditional performance of sorting green coffee beans.

Citation:

LUMAGUI, K. N., MANUEL, L. J., QUILLOY, E., & YAPTENCO, K. (2020). Varietal Classification of Selected Green Coffee Beans (Coffea arabica L. and Coffea canephora Pierre ex A.Froehner) Using Image Processing Software. Philippine Journal of Agricultural and Biosystems Engineering, 16(2), 29–44. https://doi.org/10.48196/016.02.2020.03