<![CDATA[This research was conducted to address the problem of the coffee bean sorting process, which is still performed manually in Empat Lawang Regency. The process is time-consuming, requires a large amount of human labor, and often results in inconsistent quality assessment. To overcome this, the study developed an automated classification system based on Support Vector Machine (SVM) utilizing Image Processing. The dataset was obtained directly from local collectors and consists of 740 coffee bean images, encompassing 286 good beans, 240 moldy beans, and 214 damaged beans. Feature extraction was performed based on three main characteristics color, size, and texture. Color features were calculated using the mean of RGB and HSV, while size features were obtained from the calculation of area, perimeter, and roundness. Texture features were extracted using the GLCM method. The SVM model was built using the RBF kernel and optimized with parameters C = 2 and gamma = 0.1. The evaluation results showed an accuracy of 94.37%, precision of 94.41%, recall of 94.37%, and an F1-score of 94.35%. The novelty of this research lies in the integration of color size texture features for the three-class classification of coffee beans using a lightweight model that is easily implementable at the MSME scale. However, the model is still limited to single-object images. Therefore, further research is suggested to include multi-bean datasets and consider deep learning methods that are more adaptive to variations in the number and position of coffee beans, such as CNN with YOLO or R-CNN.]]>
Copyrights © 2025
