69 / 2025-04-15 16:36:05

Video-based Detection for Feeding and Drinking Behavior of Fattening Pigs

Behavior Monitoring, Computer Vision, Precision Animal Husbandry, Health Management, Deep Learning

3. 智慧牧场与动物福利 > （2）智慧养殖的大数据、机器学习和物联网应用

Pig feeding and drinking behaviors are closely related to their physiological health, making accurate monitoring of these behaviors crucial for optimizing production efficiency and enabling early disease detection. However, existing multi-object tracking algorithms face challenges such as frequent occlusions, identity (ID) switches, and unstable long-term tracking in pig farming scenarios. This study employs an improved DeepSORT algorithm tailored for group-housed pigs, which we developed for precise pig behavior recognition. The detection algorithm is enhanced to detect pig drinking behavior by determining whether the vertical position of the pig head's detection box is above or below the water bowl's detection box. For feeding behavior recognition in group-housed pigs, individual segmentation followed by redetection is adopted to achieve precise individual behavior recognition. The precision of the drinking behavior detection algorithm averages at 95.06%, with an average recall rate of 95.19%, while the feeding behavior detection algorithm achieves an average precision of 98.48% and an average recall rate of 99.41%. Statistical analysis of feeding and drinking behaviors reveals that pigs with abnormal weight gain exhibit significantly lower feeding and drinking durations (22.2% and 11.2% lower, respectively) compared to normal pigs. Additionally, a correlation between feeding and drinking behaviors is observed. This study not only provides valuable insights for computer vision-based group-housed pig behavior recognition but also aids in physiological health assessments and remote diagnosis. It potentially serves as a reference for comprehensive feeding monitoring in group-housed finishing pigs and holds constructive value for predicting pig health through feeding and drinking behaviors.