26 / 2025-03-30 19:37:20

Automatic Early Detection of Lameness in Sows Based on Kinematics Analysis Using DeepLabCut

Lameness Detection,Sows,DeepLabCut,Kinematics Analysis,Automated Detection,Animal Welfare

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

Abstract: Lameness is a common health issue in sow farming that affects the animals’ mobility and welfare. Early signs of lameness are difficult to detect promptly due to the insufficient accuracy and reliability of traditional detection methods. To achieve the automated and accurate early detection of lameness in sows, this study collected the video data of group-housed sows in pens from both top and side views. A keypoint detection model was developed based on the DeepLabCut deep learning framework to automatically label and track multiple anatomical keypoints on the sow’s body, including 3 keypoints on the back, 3 joints on the front legs and 3 joints on the hind legs. Through the key frame identification and time-series analysis of the keypoint trajectories, kinematic features including front hoof lift-off angle, back arch angle, and support phase were extracted, and a support vector machine (SVM) classifier was used to identify lameness from these features. Experiments on 50 sows were conducted, with the data divided into training and test sets, and the model’s performance was evaluated using ten-fold cross-validation. The results showed that the proposed method achieved an average lameness detection accuracy of over 92%. Furthermore, the method demonstrated good robustness under different camera views.