Proximal sensing of barley resistance to powdery mildew

The identification and characterization of resistance and susceptibility of crop plants to fungal pathogens are essential for selecting resistant genotypes. In breeding practice, phenotyping of genotypes is realized by time consuming and expensive visual plant ratings. Resistance reactions of plants and disease pathogenesis are linked to structural and biochemical changes, which may affect the reflectance spectra of plants. In this study, hyperspectral proximal imaging in combination with data mining methods were used as a non-invasive sensor technique to assess changes in the reflection spectrum of barley differing in resistance to powdery mildew. Data mining methods can facilitate the analysis of hyperspectral images, i.e. to overcome the need for costly and time consuming manual interventions. Since plant phenotyping regularly confronts the problem of dealing with massive, high dimensional and temporal observations, data-driven techniques were employed. These data driven techniques have already been shown to be successful for mining biotic and abiotic stress. This approach meets the challenge in scalability and also provides interpretable results and models, and is easy to understand. This will be the basis for hyperspectral phenotyping of resistance reactions of crop plants in order to accelerate and automate phenotyping in resistance breeding.