Integrating UAS, Computer Vision and AI for Targeted Management of Invasive Insect Pests in Vineyards

Abstract. The increasing globalisation of trade and climate change are accelerating the spread of invasive pest species, posing significant threats to agriculture. The Japanese beetle (Popillia japonica Newman), first recorded in northern Italy in 2014, is a highly destructive pest with severe economic impacts, particularly in viticulture. Effective monitoring is essential for timely intervention, yet conventional field-based surveys are resource-intensive and limited in spatial coverage.
This study presents a novel UAV-based monitoring framework integrating near-infrared (NIR) imaging and machine learning algorithms to detect Popillia japonica adults in vineyard environments. Field experiments were conducted in two commercial vineyards in northern Italy during the beetle's summer flight season. A standardised and replicable aerial data acquisition protocol was developed using lightweight multispectral sensors mounted on rotary-wing UAV platforms. Detected insect signatures were processed through a custom CV pipeline and validated through entomological ground truthing via manual counts.
Results show a strong correlation between CV-derived detections and manual observations, with Pearson correlation coefficients ranging from 0.89 to 0.96. Although the system tends to overestimate insect counts under certain canopy conditions slightly, its integration into a GIS environment enabled the near real-time generation of prescription maps. These maps were used to guide site-specific drone spraying treatments, applying insecticides only in hotspot areas where infestation thresholds were exceeded.
This UAV-enabled, semi-automated monitoring approach significantly reduces survey time and human exposure to agrochemicals, while supporting precision pest management at scale. The methodology offers a promising framework for integrating remote sensing, AI, and entomological validation, with broader applications for managing invasive species in precision agriculture contexts.