ABSTRACT
Nowadays livestock management is critical for a country’s economy, which includes identification of breed, total count of cattle in a region, and identification of unique cattle. The livestock management is very important for the government, when insurance claims are made during floods or epidemic. Hence advanced techniques are required to use biometrics like muzzle images to uniquely identify the cattle. The cattle identification system’s aim is to identify individual cattle with its unique muzzle print. Similar to finger print of human being, every individual cattle possess unique muzzle patterns. With the feature extraction techniques, the unique extracted features could be matched against the template of cattle to identify it. The feature matching based on YOLO V5 algorithm has obtained an accuracy of 80%, whereas the system that uses feature extraction and deep learning methodology like SIFT, CNN and VGG16 trained with more than 200 cattle images has obtained accuracy of 95%. The extracted features are stored and could be used for matching and identifying the cattle in future. This would prevent many issues like false insurance claim, help the abattoir to track their cattle, etc.
