Honey bees are essential pollinators and ecological indicators, yet conventional monitoring approaches remain labor-intensive and limit the assessment of colony dynamics across biologically relevant time scales. Hive entrance activity, defined as the movement of bees entering and leaving the hive, serves as a direct proxy for foraging effort, resource availability, and colony condition. However, manual observation is restricted in duration and resolution, constraining continuous evaluation of colony responses to environmental change. In this study, a non-invasive automated monitoring framework was developed to quantify hive entrance activity under diverse environmental conditions. The system integrates image-based detection, multi-object tracking, and direction-aware trajectory analysis to identify and quantify incoming and outgoing bees. Trained on 12,013 annotated images representing varied lighting conditions, backgrounds, and activity levels, the framework achieved high detection accuracy (mAP@50 = 0.96–0.98) and real-time processing capability. Video-based evaluations showed that frame-level detections can be translated into entrance and exit estimates, although counting accuracy varied with traffic density and occlusion. These findings demonstrate that automated monitoring enables continuous measurement of biologically meaningful indicators of colony dynamics. The proposed approach provides a scalable tool for investigating foraging activity, seasonal patterns, and colony responses to environmental stressors, supporting improved ecological assessment and pollinator conservation.
Cansu Özge Tozkar (Wed,) studied this question.