Detection of airborne pathogenic wheat rust spores using machine-learning-assisted optical imaging

An advanced and efficient strategy to mitigate the impact of fungal diseases on crops focuses on the improvement of disease monitoring and fungal spore tracking. This allows for a fast localization of fungal spores even before disease symptoms can be observed by other methods. We introduce a simple and compact optical imaging system in combination with a machine-learning-assisted spore identification algorithm to detect airborne wheat rust spores. The optical imaging system is designed for the application in agricultural fields and is optimized to have a large field of view while ensuring sufficient optical resolution and keeping the system simple and compact. The imaging unit collects images under different illumination conditions supplementing the list of characteristic properties by luminescence data to improve the spore identification. The images are processed by applying multiple masks and filters and the resulting spore candidates identified with high accuracy by a machine-learning-assisted algorithm.