Behavior Tracking and Analyses of Group-Housed Pigs Based on Improved ByteTrack

Area of Science:

• Agricultural Science
• Computer Vision
• Animal Science

Background:

• Daily behavioral analysis of group-housed pigs is crucial for early detection of health problems and welfare concerns in smart pig farming.
• Automated monitoring systems are needed to efficiently analyze pig behavior and provide timely insights.

Purpose of the Study:

• To develop an automated method, Pig-ByteTrack, for monitoring and analyzing group-housed pig behavior.
• To detect health problems and improve animal welfare promptly using advanced computer vision techniques.

Main Methods:

• The Pig-ByteTrack method incorporates target detection, Multi-Object Tracking (MOT), and behavioral time computation.
• The YOLOX-X detection model was used for pig detection and behavior recognition, followed by Pig-ByteTrack for tracking.
• Performance was evaluated using 1-minute and 10-minute video datasets, measuring metrics like Higher Order Tracking Accuracy (HOTA) and Multi-Object Tracking Accuracy (MOTA).

Main Results:

• Pig-ByteTrack achieved high accuracy in 1-minute videos: 72.9% HOTA, 91.7% MOTA, 89.0% IDF1, and 41 ID switches.
• Significant improvements were observed compared to existing methods like ByteTrack and TransTrack.
• In 10-minute videos, Pig-ByteTrack achieved 59.3% HOTA, 89.6% MOTA, 53.0% IDF1, and 198 ID switches.

Conclusions:

• The Pig-ByteTrack method demonstrates efficacy in pig behavior recognition and tracking.
• This technology offers valuable technical support for the health and welfare monitoring of pig herds in smart farming environments.
• Automated behavioral analysis is key to enhancing animal welfare and farm management.