Detection of Airborne Pathogenic Wheat Rust Spores Using Machine-Learning-Assisted Optical Imaging

Fast localization and monitoring of airborne pathogens, such as fungal spores, are crucial for efficient crop disease management. Wheat rust fungi represent a major threat, as their urediniospores disperse rapidly via wind or raindrops, causing severe crop damage and yield losses. Given that wheat is the most extensively cultivated crop worldwide, outbreaks of rust diseases pose a significant risk to global food security. In this work, we present a compact optical imaging platform integrated with a machine-learning-based classification algorithm, forming an autonomous sentinel unit for in-field detection and identification of airborne urediniospores of wheat rusts. This automated device collects multichannel images of airborne particles under different illumination conditions, including a luminescence channel, and processes them using a Bayesian algorithm for fast image segmentation and spore identification within minutes and achieves an F$_1$ score of 97.7% for spore detection and 91.6% for identifying wheat rust spores. Using this system, wheat rust diseases can be localized in their early development stages, and preventative control strategies deployed even before the first symptoms become visible.