Analysis of RGB Images to Identify Local Lesions in Rosa sp. cv. Brighton Leaflets Caused by Sphaerotheca Pannosa in Laboratory Conditions

Fresh cut rose flowers are an item of considerable economic importance in Colombia. However, factors such as pests and diseases cause significant yield losses. Powdery mildew (Sphaerotheca pannosa) can cause losses of up to 30% of the total production. Thence, it is necessary to implement techniques that allow an early and objective symptoms detection in crop fields to take accurate decisions. The use of RGB and multispectral images for the detection of pests and diseases have been successfully consolidated in several agroecosystems with detection capabilities of up to 95%. In a commercial rose crop in the Bogota plateau, leaflets with and without the presence of visible signs of S. pannosa were collected from rose plants cv. “Brighton”, then, in laboratory conditions, leaflets images were acquired, pre-processed, and transformed to HSV color space. Three classes corresponding to the region of interest extracted from no visible lesion (NL), visible lesion (L), and adjacent to the lesioned area (AL) for each leaflet were compared to determine pixel intensities differences. A Welch’s ANOVA indicated a significant difference (P < 0.0001) in the leaflet reflectance between leaf tissue samples with or without visible signs of lesions caused by S. pannosa. Further experiments in field conditions that can contribute to understanding the reflectance profiles in leaflets affected by S. pannosa can be conducted to provide diagnostic tools to make more accurate decisions.