Machine learning and computer vision techniques in continuous beehive monitoring applications: A survey

Abstract

The wide use and availability of machine learning and computer vision techniques allows developing relatively complex monitoring systems in multiple domains. Besides the traditional industrial segments, new applications appear not only in biology and agriculture, where they may be employed to detect infection, parasites, and weeds, but also in automated monitoring and early warning systems. This trend clearly reflects the introduction of easily accessible hardware and development kits, such as the Arduino or RaspberryPi family. In this article, more than 50 research projects focusing on automated beehive monitoring methods using computer vision procedures are referenced; most of the approaches then facilitate pollen and Varroa mite detection together with bee traffic monitoring. Such systems could also find use in monitoring and inspecting the health state of honeybee colonies, exhibiting a potential for identifying dangerous conditions before the situation becomes critical and improving periodical bee colony inspection planning to markedly reduce the costs. By extension, our article proposes an analysis of the research trends in the given application field and outlines possible development directions. The entire project has also targeted veterinary and apidology professionals and experts, who might benefit from a matter-of-fact interpretation of machine learning and its capabilities; thus, each family of techniques is preceded by a brief theoretical introduction and motivation related to the relevant base method. The article can inspire other researchers to employ machine learning techniques in specific beehive monitoring applications.