Apple powdery mildew infestation detection and mapping using high-resolution visible and multispectral aerial imaging technique

Abstract

Powdery mildew (PM) in apple orchards is a critical fungal disease that considerably reduces yield, harvested fruit quality, and orchard health. Rapid detection and mapping of resulting infestation at orchard scale is a challenge with existing laborious manual scouting approaches. Therefore, this study explored the feasibility of detecting and mapping PM infestation in an apple orchard block using high-resolution visible (red-green-blue [RGB]) and multispectral imaging technique. Imaging campaigns were conducted over an experimental orchard using small unmanned aerial systems (UAS) integrated with the above optical sensors. K-means classifier trained on individual snapshots of RGB imagery had mean disease detection accuracy of 77%. Eight multispectral vegetation indices also showed significant differences (p < 0.001) between healthy (Mean: 0.25–0.84) and infected (Mean: 0.01–0.25) leaves. Modified Simple Ratio-Red (MSRR), Modified Simple Ratio-Blue (MSRB), and Optimized Soil Adjusted Vegetation Index (OSAVI) showed the highest contrast (Mean: 0.46–0.79). Orchard block-scale mapping using RGB orthomosaic layer, classified with k-means and spectral angle mapper techniques had accuracies up to 73%. Overall, high resolution imagery in the visible and multispectral domain could sufficiently detect the infection and generate site-specific disease heat maps. These maps could be useful to apple growers in directing prescriptive resources ( pruning labor or fungicide application) for PM management.