Holstein Cow Identification By Black And White Pattern Using KL Expansion

Abstract

Holstein cow identification is very important from points of pedigree registration, artificial insemination, livestock hygiene and food safety control. Earmark identification system is often used in Western countries and Japan. In this study, a black and white pattern identification system using image processing was investigated as an easier and more effective identification system. 20 Holstein cows could be identified perfectly using Euclid distance for their patterns matching in an eigen space which was constructed by Karhunen-Loeve expansion, however, assuming that cow's posture and illumination was invariable in this experiment. After two degree counterclockwise rotation of cow's posture in a computer, their cows could not be identified perfectly. The identification algorithm must be improved to be robust to the relative posture change between cow and camera.