A Beef and Pork Meat Image Identification System Using Hybrid Feature Extraction and Support Vector Machine

Abstrak

Beef and pork differentiation is an important issue in halal food authentication, particularly in Muslim-majority countries such as Indonesia, where meat adulteration can reduce consumer trust and violate religious dietary requirements. Since raw beef and pork often exhibit similar visual characteristics, manual inspection is frequently unreliable. This study proposes an automated classification system for beef (halal) and pork (non-halal) using Support Vector Machine (SVM) with hybrid features that combine statistical color descriptors (mean, variance, skewness, kurtosis) and texture descriptors (hausdorff dimensions). The results demonstrated that both the Linear and Polynomial kernels achieved the highest accuracy of 98,08%, while the RBF kernel achieved 90,38%. Further comparison with K-Nearest Neighbor and Naïve Bayes classifiers showed that SVM with a Linear kernel outperformed the other methods, with KNN and Naïve Bayes achieving 90,38% and 96,15% accuracy, respectively. These findings indicate that SVM with a Linear kernel is highly effective for differentiating beef and pork, offering a reliable tool to support halal authentication and consumer protection.

Kata kunci

• Algoritma Support Vector Machine
• Beef
• feature extraction
• halal
• pork