Feature Extraction and Automatic Recognition Model Construction for Head Back Posture During the Parturition Process...

Area of Science:

• Veterinary Science
• Animal Behavior
• Computer Vision

Background:

• The 'head back' posture during bovine parturition is a key indicator of pain and labor.
• Real-time monitoring of calving behaviors can improve animal welfare and facilitate timely assistance.

Purpose of the Study:

• To develop and evaluate a computer vision model for recognizing head posture during parturition.
• To analyze behavioral differences in primiparous versus multiparous cows during calving.

Main Methods:

• Utilized the YOLOv8 algorithm trained on 25,617 images to classify cow postures.
• Assessed model performance across various camera angles and environmental conditions (lighting, time of day).
• Conducted behavioral analysis comparing primiparous and multiparous cows.

Main Results:

• The YOLOv8 model achieved a 0.71 F1 score for posture recognition.
• Posture recognition accuracy exceeded 90% from abdominal camera angles and remained stable under diverse lighting.
• Primiparous cows showed significantly longer parturition durations and head back posture durations.

Conclusions:

• Computer vision offers a promising tool for real-time monitoring of bovine parturition.
• The developed model can aid in identifying calving difficulties and enhancing animal welfare.
• Significant behavioral differences exist between primiparous and multiparous cows during calving.