Optimizing black cattle tracking in complex open ranch environments using YOLOv8 embedded multi-camera system.

Area of Science:

• Agricultural Engineering
• Computer Vision
• Animal Science

Background:

• Daily activity monitoring is vital for black cattle well-being and farm management.
• Advancements in AI and computer vision offer new possibilities for livestock monitoring.
• Traditional single-camera systems struggle with accurate tracking in complex open ranch environments.

Purpose of the Study:

• To develop a novel Customized Multi-Camera Multi-Cattle Tracking (MCMCT) system for precise monitoring of black cattle.
• To overcome the limitations of single-camera setups in detecting and tracking cattle in open ranch environments.
• To enhance livestock behavior analysis, health evaluation, and estrus event prediction in precision farming.

Main Methods:

• The MCMCT system utilizes a tracking-by-detection model enhanced with the YOLO v8 segmentation model as the backbone.
• A multi-camera approach (four cameras) is employed to improve coverage and accuracy in complex environments.
• Integration with the Segment Anything Model (SAM) automates cattle mask region extraction, reducing manual labeling.

Main Results:

• The YOLOv8-MCMCT system achieved an average Multi-Object Tracking Accuracy (MOTA) of 95.61% across 10 cases using 4 cameras.
• The system processed data at 30 frames per second, demonstrating efficient real-time performance.
• Comparative analysis showed superior performance of MCMCT over state-of-the-art tracking methods like Bot-sort, Byte-track, and OC-sort.

Conclusions:

• The MCMCT system provides a robust, non-contact solution for automatic livestock monitoring in precision cattle farming.
• The system demonstrates high accuracy and adaptability across varied ranch environments without extensive retraining.
• This research significantly contributes to improving livestock monitoring, agricultural efficiency, and farm management practices.