Precision Phenotyping of Wheat Lodging Using UAVs

Crop lodging poses significant challenges for cereal crops like wheat and rice, leading to difficulties in harvesting and reductions in both yield and quality. Therefore, identifying and developing lodging-resistant varieties is crucial in breeding programs. Phenotyping plays a critical role in these breeding programs to accelerate genetic gain. Recent advancements in sensor technology, platforms, and modeling methods have enabled high-throughput phenotyping of various traits across different crop varieties. In this study, we present a method for phenotyping wheat lodging usings RGB cameras mounted on Unmanned Aerial Vehicles (UAVs). We used U-Net, a semantic segmentation-based approach, to segment lodged pixels in RGB images of wheat research plots captured by UAVs, thereby estimating the percentage of lodging in each plot. Comparisons between the model’s estimated lodging percentages and manual observations at multiple locations indicate that this method accurately estimates the lodging phenotype. The average BLUE (Best Linear Unbiased Estimator) correlation and heritability score across 6 locations were 0.8 and 0.45 respectively.