Aedes albopictus (Skuse), an invasive disease vector, poses a nuisance and public health threat to communities in the Northeastern United States. Climate change and ongoing adaptation are leading to range expansion of this mosquito into upstate New York and other northeastern states. Organized mosquito control can suppress populations, but it is time consuming, costly, and difficult as Ae. albopictus oviposits in small, artificial, water-holding containers. Unmanned aerial vehicles (UAVs), with centimeter-resolution imaging capabilities, can aid surveillance efforts. In this work, we show that a convolutional neural network trained on images from a UAV is able to detect Ae. albopictus habitat in suburban communities, and the number of containers successfully imaged by UAV predicted the number of containers positive for mosquito larvae per home. The neural network was able to identify some, but not all, potential habitat, with up to 67% precision and 40% recall, and can classify whole properties as positive or negative for larvae 80% of the time. This combined approach of UAV imaging and neutral network analysis has the potential to dramatically increase capacity for surveillance, increasing the reach and reducing the time necessary for conventional on-the-ground surveillance methods.
Keywords: Aedes albopictus; aerial surveillance; machine learning; vector control.
© The Author(s) 2020. Published by Oxford University Press on behalf of Entomological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.