Automatic Cattle Identification based on Muzzle Photo Using Speed-Up Robust Features Approach

Cattle identification has been a serious problem for breeding association. The need of a robust identification method is a must. The previous identification means have not been satisfactory. The biometric marking has been investigated to be a permanent marking of the individual. Muzzle pattern or nose print has the same characteristic with the human fingerprint which is the most popular biometric marker. SURF approach which is an object recognition based method has been evaluated for the automatic cattle identification purpose. Based on the experiment result SURF approach outperforms the previous research that is used eigenface algorithm. The original SURF approach relatively can handle non-normalized data set (scale and orientation invariant) with high accuracy and precision. With a sufficient training data, the performance of the original SURF can be more than 0.9 in accuracy and kappa statistic. The U-SURF as another version of the original SURF has shown an outstanding performance more than the original SURF and eigenface algorithm, but only in the orientation normalized data. Key–Words: Animal biometric, Cattle identification, Cattle’s Muzzle photo, SURF, U-SURF.