Automated lab identification and sorting (ALIDAS) for mosquito surveillance

抽象的。蚊子每年造成近50万人死亡。蚊虫 在过去的几年里，病媒控制努力减轻了蚊媒疾病的负担 几十年，但数据驱动的病媒控制决策方面的限制阻碍了进展。 蚊子监测-监测一个地区以了解蚊子的种类组成， 丰度和空间分布--使蚊子控制组织能够有效、 高效、明智的蚊子控制决策。尽管蚊子的种类很重要 鉴定在监视中，形态鉴定仍然具有高度的资源、时间和 劳动密集型。雇佣季节性员工，这是蚊子控制的一项重大经常性成本 组织，是扩大能力的常规做法。昆虫学专业知识也可以 根据个人训练和经验差异很大，并导致不正确的物种 身份证明。我们寻求开发第一个自动化实验室识别和分类(ALIDAS) 蚊媒监测系统，以提高监测能力并及时生成 用于定向控制蚊子媒介的数据。计算机视觉具有规模识别的潜力 不同种类的蚊子；然而，自动化昆虫学实验室工作流以 最大限度地节省运营成本，需要系统地处理蚊子和 运动，同时保存和捕获图像中的诊断形态特征 分类。这项提议将利用新的光学、气动和计算机视觉技术 利用计算机视觉分离、处理和识别物种蚊子标本的方法。 最终，这里开发的方法将使蚊子控制组织能够利用 实用系统中的图像识别将提高昆虫学实验室的能力和 容量，同时降低运营成本。

Abstract. Mosquitoes are responsible for nearly half-a-million deaths each year. Mosquito vector control efforts have reduced the burden of mosquito-borne diseases over the past several decades, but limitations in data-driven vector control decision making hinders progress. Mosquito surveillance—monitoring an area to understand mosquito species composition, abundance, and spatial distribution—enables mosquito control organizations to make effective, efficient, and judicious mosquito control decisions. Despite the importance of mosquito species identification in surveillance, morphological identification remains highly resource, time, and labor intensive. Hiring seasonal staff, a significant recurring cost to mosquito control organizations, is the conventional practice to expand capacity. Entomological expertise can also vary widely based on individual training and experience, and result in incorrect species identifications. We seek to develop the first automated lab identification and sorting (ALIDAS) system for mosquito vector surveillance, to increase surveillance capacity and generate timely data for targeted mosquito vector control. Computer vision has potential to scale identification across diverse mosquito species; however, automating the entomological lab workflow to maximize operational savings, requires a systematic approach to mosquito handling and movement, while preserving and capturing diagnostic morphological characters in images for classification. This proposal will utilize novel optics, pneumatics, and computer vision approaches to isolate, handle, and identify mosquito specimens to species with computer vision. Ultimately the approaches developed here will allow mosquito control organizations to leverage image recognition in a practical system that will increase entomological lab capability and capacity, while reducing operational costs.