Recognition of defect types of Arabica coffee beans using image processing

Coffee is one of the most popular commodities in the world and Indonesia is a large coffee producer. Every year coffee production in Indonesia reaches more than 600,000 tons and some of them are exported. However, most coffee bean grading is still done manually, and this work requires a lot of time with low consistency. This study aims to study the quality parameters of Arabica coffee beans and the use of image processing to distinguish types of defects in green coffee beans. Samples were grouped into normal beans, shelled beans, deformed, discolored, hollow beans and small beans. Each group consisted of 100 green coffee bean samples. Image acquisition is carried out on white paper from a short distance using USB camera. Image processing program is made to obtain several image parameters from objects such as area, perimeter, height and width of beans and color values. Recognition results using image processing sequentially were 77%, 43%, 41%, 90%, 68%, and 84% for normal beans hollow beans, deformed beans, discolored beans, skinned beans, and small beans.