Vision-Based Smart Beehive Integrating Environmental Sensing for Climate-Smart Apiculture
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
This study presents a field-deployable smart beehive that combines real-time, vision- based in/out counting with continuous environmental sensing on low-power edge hardware. A four-zone entrance model with a conservation constraint (ΔA = 0) is introduced to regularize event inference, and practical stabilizers—a boundary buffer, short debouncing, and per-track cooldown—are used to suppress spurious transitions caused by jitter, occlusion, and brief reversals on the entrance lip. The system runs on a Raspberry Pi with an optional Edge TPU accelerator and exposes a dashboard that collocates entrance activity with internal and external temperature, humidity, CO2, and hive weight for management decisions. Field experiments at two outdoor sites show that the proposed Algorithm 2 achieves event-level IN accuracy of 82.8–88.2% and OUT accuracy of 97.1–100.0%, substantially reducing double counts relative to an aggregate-zone baseline while maintaining real-time performance on embedded hardware. By interpreting the joint activity and microclimate time series in the context of climate-smart agriculture, precision apiculture, and agricultural- environment indicators, we demonstrate how the smart beehive can function not only as an automated counter, but also as a sensor node for pollinator-aware, climate-smart farm management.
Abstract
This study presents a field-deployable smart beehive that combines real-time, vision- based in/out counting with continuous environmental sensing on low-power edge hardware. A four-zone entrance model with a conservation constraint (ΔA = 0) is introduced to regularize event inference, and practical stabilizers—a boundary buffer, short debouncing, and per-track cooldown—are used to suppress spurious transitions caused by jitter, occlusion, and brief reversals on the entrance lip. The system runs on a Raspberry Pi with an optional Edge TPU accelerator and exposes a dashboard that collocates entrance activity with internal and external temperature, humidity, CO2, and hive weight for management decisions. Field experiments at two outdoor sites show that the proposed Algorithm 2 achieves event-level IN accuracy of 82.8–88.2% and OUT accuracy of 97.1–100.0%, substantially reducing double counts relative to an aggregate-zone baseline while maintaining real-time performance on embedded hardware. By interpreting the joint activity and microclimate time series in the context of climate-smart agriculture, precision apiculture, and agricultural- environment indicators, we demonstrate how the smart beehive can function not only as an automated counter, but also as a sensor node for pollinator-aware, climate-smart farm management.