Automated evaluation of dairy goat body condition and analysis of differences in lumbar features, milk composition and blood biochemical indicators

Abstract

Body condition scoring (BCS) in dairy goats serves as an objective method for quantifying the reserves of body tissues, namely adipose and muscle tissues, and evaluating overall health. Recognized as a dependable and feasible welfare indicator, BCS is crucial for managing animal health. This study introduced an automated BCS model for dairy goats, leveraging computer vision and deep learning via the YOLO v5 algorithm. The model distinguished waist phenotypic characteristics, analyzed milk quality, and assessed blood biochemical indices across different body conditions. Demonstrating high precision, the model achieved Precision (P), Recall (R), and F1 scores of 78.5 %, 82.0 % and 81.7 %, respectively. It effectively identified underweight, moderate, and overweight groups with identification rates of 85.2 %, 79 %and 71.2 % respectively, and maintained a deviation rate from manual assessments of ≤ 10 %. Notably, the waist region's grayscale parameters and brightness levels correlated positively with body condition scores, while the depth of indentation exhibited a negative correlation. Milk yield showed no significant variation (3–4 kg), but milk protein content was highest in the average condition group. Lipid and liver metabolism markers varied significantly with body condition, underscoring physiological impacts. This model not only confirmed the robustness of YOLO v5 for animal welfare assessment but also early intervention strategies are used in the management of dairy goats, in line with the principles of precision animal husbandry, particularly beneficial during the critical mid-lactation period. This work underscored the significant potential of integrating advanced technologies into everyday agricultural practices to enhance animal welfare and farm management.