Shinta Widyaningtyas, Muhammad Arwani, Sucipto, Yusuf Hendrawan
Palm civet coffee, a highly prized specialty beverage, is increasingly targeted by adulteration due to its limited production and substantial market value. The absence of reliable detection methods necessitates the development of efficient, non-destructive sensing techniques. The aims of this study is a development of machine vision in selecting relevant color and texture features in detecting the purity of civet coffee. The development of computer vision is expected to be more applicable in the creation of Palm Civet Coffee purity detection tools. A comprehensive dataset comprising 528 coffee bean images was extracted and generated 101 image features (11 color and 90 textural). Utilizing a wrapper method, three classifiers k-Nearest Neighbors (kNN), Random Forest (RF), and Support Vector Machine (SVM) were coupled with four optimization algorithms (Bat Algorithm, Cuckoo Search, Genetic Algorithm, and Grey Wolf Optimizer) to identify the most informative features. Comparative performance analysis revealed that the Random Forest classifier, optimized with the Grey Wolf Optimizer achieved the highest accuracy of 0.981 using five features i.e Blue_Mean, Hue_Entropy, Gray_Inverse, S_HSL_Cor-relation, and Green_Cluster. On the other hand, kNN and SVM classifiers attained maximum accuracies of 0.943 and 0.925, respectively. Additionally, statistical analysis using One-Way ANOVA demonstrated significant performance improvements of the accuracy for each classifier (P-value < 0.01). These findings demonstrate the efficacy of the wrapper method in feature selection for classifying purity of Palm Civet Coffee. This research is one of the stages in the development of automatic civet coffee purity detection to prevent adulteration so that authenticity is guaranteed. © 2025, Editora UFLA. All rights reserved.
Agriculture Engineering, Politeknik Negeri Jember, East Java, Jember, Indonesia; Agroindustrial Technology, Universitas Nahdlatul Ulama Indonesia, Central Jakarta, Jakarta, Indonesia; Agroindustrial Technology, Universitas Brawijaya, East Java, Malang, Indonesia; Biosystem Engineering, Universitas Brawijaya, East Java, Malang, Indonesia
