Method for Dairy Cow Target Detection and Tracking Based on Lightweight YOLO v11

Area of Science:

• Precision Livestock Farming
• Computer Vision
• Animal Welfare

Background:

• Effective dairy cow monitoring is crucial for precision livestock farming, aiming to enhance animal health and welfare.
• Existing methods face challenges with large model parameters, inaccurate multi-target tracking, and complex nonlinear cow movements.

Purpose of the Study:

• To develop a lightweight object detection model for dairy cow motion monitoring.
• To compare the performance of different tracking algorithms for improved accuracy.

Main Methods:

• An improved YOLOv11n model incorporating the Ghost module and ELA attention mechanism was developed.
• A novel SDIoU loss function was implemented to address varying cow target sizes.
• Four tracking algorithms (ByteTrack, BoT-SORT, OC-SORT, BoostTrack) were evaluated.

Main Results:

• The improved YOLOv11n model achieved an 18.59% reduction in parameters, with increased mAP@75 (2.0%) and mAP@50-95 (2.3%).
• The OC-SORT algorithm demonstrated superior performance, achieving 97.02% MOTA and 89.81% HOTA.
• The combined approach offers a lightweight yet high-performance solution for cow tracking.

Conclusions:

• The proposed lightweight object detection model significantly reduces computational load while maintaining high accuracy.
• The OC-SORT algorithm provides robust cow tracking capabilities using standard video surveillance.
• This research facilitates continuous cow monitoring, supporting improved dairy farm management and animal welfare.