Precision Beekeeping and Selective Breeding of the Honey Bee Using Multimodal Signal Processing and Machine Learning

虽然蜜蜂主要与蜂蜜生产有关，但作为授粉者，蜜蜂在农业中发挥着更重要的作用。事实上，世界上近三分之一的食物直接依赖蜜蜂授粉。在加拿大，养蜂业是一个重要的农业产业，生产蜜蜂用于销售、蜂蜜和其他蜂箱产品，以及提供有价值的授粉服务。然而，在过去的十年里，蜂群以惊人的速度死亡。已经确定的两个主要因素包括传播致命病毒的瓦螨寄生虫的增殖和蜂王的失败。蜂巢的目视检查是目前蜜蜂健康监测的现有做法。然而，打开蜂箱会破坏殖民地，扰乱蜜蜂的行为，并限制可以做的事情的范围。为了克服这一限制，Nectar Technologies Inc.开发了基于传感器的技术，这些技术被放置在蜂箱内，可以同时对几个蜂箱的健康状况进行远程评估。到目前为止，他们的解决方案依赖于温度和湿度传感器来监测蜂王的存在，并依赖于外部秤来监测蜂巢重量，这是蜂巢健康和蜂蜜产量的相关因素。该项目旨在将这个“蜂巢互联网”平台提升到一个新的水平。通过与德尚博动物科学研究中心（Centre de recherche en sciences animales de Deschambault）（CRSAD，一种商业蜂王饲养器），并在蜂箱内集成额外的振动、声学和CO2传感器，我们建议开发五项关键创新：1）建立表型群体性能性状的多模态相关性以允许远程蜂王性能监测和2）用于选择最佳性能群体的遗传参数，3）开发自动化方法来测量菌落活性和强度，4）蜂蜜生产和流动，以及5）蜂巢健康。拟议的系统具有低成本、耐用和可靠的潜力，因此将可供所有加拿大养蜂人使用，无论规模大小。拟议的项目还将允许发现蜜蜂动力学和健康监测背后的新科学。因此，预计将为加拿大带来重大的社会、科学和经济效益。

While mostly associated with honey production, honeybees play a more crucial role in agriculture as pollinators. In fact, nearly one third of the world's food is directly dependent on bee pollination. In Canada, beekeeping is an important agricultural industry, with bees being produced for sale, honey and other hive products, as well as delivering valuable pollination services. Over the last decade, however, bee colonies have been dying off at alarming rates. Two major factors that have been identified include the proliferation of the varroa destructor parasite, which spreads deadly viruses, and failing queen bees. Visual inspection of hives is the current existing practice of bee health monitoring. Opening a beehive, however, aggravates the colony, disrupts honeybee behavior, and limits the scope of what can be done. To overcome this limitation, Nectar Technologies Inc. has developed sensor-based technologies that are placed within hives and allow for remote assessment of the health of several hives simultaneously. To date, their solutions have relied on temperature and humidity sensors to monitor queen bee presence and on external scales to monitor hive weight, a correlate of hive health and honey production. This project aims to take this "Internet of Hives" platform to the next level. By partnering with the Centre de recherche en sciences animales de Deschambault (CRSAD, a commercial queen bee breeder) and integrating additional vibration, acoustic, and CO2 sensors within the hives, we propose to develop five key innovations: 1) build multimodal correlates of phenotypic colony performance traits to allow for remote queen performance monitoring and 2) genetic parameters to be used to select best performing colonies, 3) develop automated methods to measure colony activity and strength, 4) honey production and flow, and 5) hive health. The proposed system has the potential to be low-cost, durable, and reliable, thus will be accessible to all Canadian beekeepers, large or small. The proposed project will also allow for the discovery of new science behind honey bee dynamics and health monitoring. As such, significant societal, scientific, and economic benefits to Canada are expected.