Klasifikasi Biji Kopi Berdasarkan Bentuk Menggunakan Image Processing dan K-NN

Abstract

Temanggung Regency is the largest coffee producing area in Central Java. Robusta and Arabica are two types of coffee grown in this area. The manual method of sorting coffee beans is still used so the results are more subjective. Therefore, we require a system for categorizing coffee beans so that the results are more objective and reliable. This study uses K-NN classification and morphological features to recognize coffee beans based on the type and shape of the coffee bean defects. This aim of this study discovers which characteristics are better at classifying coffee beans into four categories (whole Robusta, whole Arabica, Robusta broken, and broken Arabica). A total of 110 coffee bean photos were used, with 80 images as training data, 40 images as test data, and a total of five morphological features.

Our findings reveal that morphological traits may classify coffee beans into four categories with an accuracy of 62.5%, which is very good for detecting 100% whole Robusta and 90% Arabica but remains weak for recognizing broken coffee beans by type. Lean performs better in distinguishing coffee beans based on four classes, with a 70% accuracy. Morphological features outperform color features in distinguishing coffee beans based on shape defects, with an accuracy score of 83%.