Foliar fungal pathogens pose significant challenges to global wheat production, necessitating the development of disease-resistant cultivars. A critical step in breeding for resistance is high-throughput phenotyping, where thousands of wheat germplasm lines are screened in field disease nurseries. Traditionally, phenotyping relies on visual disease scoring, which is labour-intensive, subjective, and prone to fatigue-induced errors, particularly in large-scale breeding programs.
This study evaluates the feasibility of using unmanned aerial vehicle (UAV)-based red, green, and blue (RGB) imagery for automated phenotyping of yellow leaf spot (YLS) disease in wheat. Several convolutional neural network (CNN) architectures were tested for their ability to classify YLS disease severity across different resistance levels. Experimental results demonstrate that UAV-RGB imagery, combined with CNN-based analysis, offers a scalable, efficient, and objective alternative to manual phenotyping. The approach shows promising accuracy in detecting disease resistance variations, with potential applications for real-time field monitoring and early intervention strategies.
Further refinement of CNN models is warranted to align disease classification with breeder-defined multi-level scoring systems. This research highlights the potential of UAV-based phenotyping in accelerating the selection of resistant wheat cultivars, thereby improving crop management and sustainability.