Machine Learning-based Prediction of Cattle Body Weight Using Muzzle Morphometrics
Swarnalata Bara
Livestock Production and Management Section, ICAR-Indian Veterinary Research Institute, Izatnagar, Bareilly -243 122, Uttar Pradesh, India.
Mukesh Singh
Livestock Production and Management Section, ICAR-Indian Veterinary Research Institute, Izatnagar, Bareilly -243 122, Uttar Pradesh, India.
Hari Om Pandey
Division of Animal Genetics and Breeding, ICAR-Indian Veterinary Research Institute, Izatnagar, Bareilly- 243 122, Uttar Pradesh, India.
Anuj Chauhan
Livestock Production and Management Section, ICAR-Indian Veterinary Research Institute, Izatnagar, Bareilly -243 122, Uttar Pradesh, India and Division of Animal Genetics and Breeding, ICAR-Indian Veterinary Research Institute, Izatnagar, Bareilly- 243 122, Uttar Pradesh, India.
Gyanendra Kumar Gaur
Krishi Bhawan, New Delhi- 110 001, India.
Ashwni Kumar Pandey
Livestock Production and Management Section, ICAR-Indian Veterinary Research Institute, Izatnagar, Bareilly -243 122, Uttar Pradesh, India and Division of Animal Genetics and Breeding, ICAR-Indian Veterinary Research Institute, Izatnagar, Bareilly- 243 122, Uttar Pradesh, India.
Ajoy Das *
Livestock Production and Management Section, ICAR-Indian Veterinary Research Institute, Izatnagar, Bareilly -243 122, Uttar Pradesh, India.
*Author to whom correspondence should be addressed.
Abstract
Body weight measurement of cattle is a tedious farm operation but is essential for their health maintenance at farm. This study proposes a novel and easier approach for cattle body weight prediction using muzzle morphometrics. Vrindavani crossbred cattle of different age groups were considered for the study. The muzzle images were collected and analyzed in MATLAB for determination of muzzle dimensions followed by mapping the dimensions to the body weight of the cattle using artificial neural network with varying network parameters. The results of the showed that all muzzle parameters had good correlation with the body weight of the cattle. Further, it was also observed that the combination of Levenberg-Marquardt training algorithm with logsigmoidal transfer function performed the best with model simulation accuracy of 78.07%. The study concludes that muzzle morphometrics may be used for body weight measurements, however, newer or diverse muzzle parameters may be considered in future works to further improve the model accuracy for a more practical application.
Keywords: Artificial neural networks, Image processing, MATLAB, muzzle prints