Regression analysis for dairy cattle body condition scoring based on dorsal images

Abstract

Body condition scoring (BCS) provides an objective assessment of the amount of subcutaneous fat appositions and energy storage of the dairy cattle, which has become a powerful tool for dairy industry management. Traditional BCS is estimated by technicians manually method involving visual and tactile aspects, which is high-cost and subjective. Hence, the, accurate and efficient BCS automatic evaluation technology is studied. In this paper, we proposed an effective dairy cattle body condition scoring method based on cow’s dorsal 2D digital images. A bounding rectangle normalization method is used to extract cow’s contour information and distance vectors are constructed to describe the shape. Then, six regression methods are discussed for BCS regression modeling. Experiment on a benchmark dataset demonstrated that elastic net obtained the best accuracy on the BCS task.