Towards rapid detection and mapping of powdery mildew in apple orchards

Powdery mildew (PM) in apples is a critical fungal disease that adversely affects yield and fruit quality. Conventional PM identification techniques are laborious. This study evaluated the suitability of non-destructive low altitude imaging in the visible (RGB) and multispectral domain for PM detection in apple orchards. Imagery snapshots were acquired at known locations of PM infestation in apple trees using two small unmanned aerial systems (UAS). One UAS employed a consumer grade RGB camera while other UAS had a five-band multispectral imaging sensor. Custom image processing algorithms were developed for RGB and multispectral imagery analysis. The RGB images were classified using k-means and resulted in PM detection accuracy of 76.4%. A strong agreement was also observed between the number of actual PM clusters and clusters that were classified (R² = 0.94). Seven vegetation indices (VIs) were extracted from the multispectral imagery as the potential chlorophyll variation indicators. Such indices showed significant difference (p <; 0.05) for PM infected leaves and the healthy leaves. Modified Simple Ratio-Red, Modified Simple Ratio-Blue and Optimized Soil Adjusted Vegetation Index were the key differentiators showing low values for PM (Mean: 0.01-0.05) and high values for healthy leaves (Mean: 0.51-0.84). Overall, results showed the potential to detect and quantify PM infestation in apple orchards at high spatial resolution. Additional efforts are needed to translate these results to map PM infestation at the orchard block level that could aid in site-specific remedial measures.