Automatic behaviour analysis system for honeybees using computer vision

Publikation: Bidrag til tidsskrift/Konferencebidrag i tidsskrift /Bidrag til avisTidsskriftartikelForskningpeer review

We present a fully automatic online video system, which is able to detect the behaviour of honeybees at the beehive entrance. Our monitoring system focuses on observing the honeybees as naturally as possible (i.e. without disturbing the honeybees). It is based on the Raspberry Pi that is a low-cost embedded computer with very limited computational resources as compared to an ordinary PC. The system succeeds in counting honeybees, identifying their position and measuring their in-and-out activity. Our algorithm uses background subtraction method to segment the images. After the segmentation stage, the methods are primarily based on statistical analysis and inference. The regression statistics (i.e. R2) of the comparisons of system predictions and manual counts are 0.987 for counting honeybees, and 0.953 and 0.888 for measuring in-activity and out-activity, respectively. The experimental results demonstrate that this system can be used as a tool to detect the behaviour of honeybees and assess their state in the beehive entrance. Besides, the result of the computation time show that the Raspberry Pi is a viable solution in such real-time video processing system.
OriginalsprogEngelsk
TidsskriftComputers and Electronics in Agriculture
Vol/bind122
NummerMarch
Sider (fra-til)10-18
Antal sider9
ISSN0168-1699
DOI
StatusUdgivet - 22 jan. 2016

    Forskningsområder

  • Video monitoring system, Raspberry Pi, Computer vision, Honeybees, Counting, In-and-out activity

Se relationer på Aarhus Universitet Citationsformater

ID: 96313929