KLASIFIKASI MUTU KUALITAS BIJI KOPI BERDASARKAN EKSTRAKSI FITUR WARNA HSV DAN TEKSTUR GLCM MENGGUNAKAN ALGORITMA SUPPORT VECTOR MACHINE

Abstract

Coffee bean quality grading across smallholder agricultural processing relies predominantly on manual visual screening, incurring substantial human error and subjectivity. This study develops an automated computer vision grading model integrating Hue-Saturation-Value (HSV) chromatic descriptors and Gray-Level Co-occurrence Matrix (GLCM) spatial texture metrics powered by a Support Vector Machine (SVM) classifier. The digital dataset consists of 600 bean samples spanning three quality grades: Grade 1 (Premium), Grade 2 (Medium), and Defect. Evaluation via 10-fold cross-validation examined Linear and Radial Basis Function (RBF) kernels. Experimental findings demonstrate that multi-domain HSV and GLCM fusion with an RBF kernel attained 100.00% classification accuracy and 100.00% F1-score, surpassing single color space feature configurations (95.83%). The framework exhibits high potential for deployment in automated conveyor sorting machines.