EAR IMAGE-BASED INDIVIDUAL PIG IDENTIFICATION BY USING STATISTICAL PARAMETERS

One of the most popular methods for addressing the shortcomings of traditional identification methods is biometric-based animal identification. The animal's phenotypic characteristics, such as its muzzle, retina, ear venation pattern, and iris, can be used in identification. The most important characteristic for pig identification is the pattern of venation in the ears. However, it is somewhat difficult to capture the entire venation web due to the thickness of ear skin. This research attempted to identify the 'Yorkshire' pig using the images of their ears. An artificial ambience was also created to block out the external noise. In present work, a total of 165 images from 5 distinct pigs were captured. The acquired images were used to collect eight statistical features. Four machine learning -based techniques, including Support Vector Machine, Decision Tree, K-Nearest Neighbor, and Random Forest, were used to assess and categorize the obtained statistical characteristics in order to forecast the specific pig. In comparison to the other approaches, Random Forest classifier showed the best identification accuracy (90.9%), followed by Support Vector Machine, Decision Tree, and K -Nearest Neighbor. The technique will help us analyze the vein patterns in pig ears to identify specific pigs.