A Robust Multiple Honeybee Tracking Method from Videos Captured at Beehive Entrance

Monitoring the health of honeybees is an important task of beekeepers to mitigate negative impacts that could happen to the beehives. In fact, the high density of honeybees at the beehive entrance and the complex movements lead to nonlinear motion and heavy occlusion challenges for honeybee detection and tracking. To tackle the aforementioned challenges, a method for honeybee detection and tracking that incorporates a real-time object detection module based on YOLOv5 and robust object tracking based on OC-SORT is proposed for honeybee detection and tracking from images captured at the beehive entrance. Extensive experiments have been conducted on a self-built dataset named VnBeeTracking. Experiment results confirm the outperformed results of the proposed method compared with other tracking-by-detection methods. The proposed method obtained 78.3% and 88.2% for two important metrics MOTA and MOTP in object tracking while maintaining a low rate of ID switch.