This study investigates the ability of Visible-Near Infrared Spectroscopy to detect powdery mildew infection in cucumbers at a symptomless stage of the disease. Experiments were performed with the cucumber variety Sandra. Plants were infected with a suspension of spores of powdery mildew (Podosphaera xanthii) in three variants - single, double and triple infection. Samples for spectral analysis were taken 3, 5, 7, 9, 11 and 13 days after the infection. Reflectance spectra of cucumber leaves were collected using a USB4000 spectrometer (OceanOptics, Inc. Dunedin, Fl., USA) with a fiber optics probe. The studied wavelength range was 450-1100nm. Two approaches were used for discrimination of between infected and healthy plants – vegetative indices used in remote sensing, and soft independent modelling of class analogy (SIMCA) classification. Differences between the reflectance spectra of the healthy plants, infected plants at a symptomless stage, and plants with developed symptoms of the disease were obtained. The largest differences were found between 540 and 680 nm, which can be explained by changes in the in the leaf structure and chemical composition of the plant tissues during pathogenesis. The calculated vegetative indices (NDVI, Red Edge Index, Photochemical Reflectance Index, Water Band Index, Carotenoid Reflectance Index) showed differences between healthy, infected plants at a symptomless stage, and diseased plants, but the differences between the healthy group and the infected plants at a symptomless stage of the disease were not statistically significant. The developed SIMCA models based on the spectral data had an accuracy rate over 78% in predicting the infection in plants. The study shows the potential of VIS-NIR spectroscopy, coupled with chemometric classification techniques, for successful distinction of a powdery mildew infection in cucumbers at the early symptomless stage of the disease.
Find this article in full text from Scitation.