Early diagnosis and pathogenesis monitoring of wheat powdery mildew caused by blumeria graminis using hyperspectral imaging

Abstract

Powdery mildew caused by blumeria graminis is responsible for wheat yield losses in combination with a decline in quality. Hyperspectral imaging as a promising non-invasive sensor technique has potential for early diagnosis and pathogenesis monitoring of wheat powdery mildew, which is a practice that allows for precision crop protection. Hyperspectral images were first captured before inoculation as healthy samples and daily 2 to 5 days after inoculation (dai) as infected ones. Principal component analysis (PCA) was applied to observe the discrimination capability between samples at different infected stages, while a gray-level co-occurrence matrix (GLCM) was used to extract textural features from the first three principal component images. Then partial least squares discriminant analysis (PLS-DA) model was developed to evaluate the ability for early diagnosis of the disease using effective wavelengths, texture features and their fusion, respectively. Compared with the models using spectral or textural feature alone, PLS-DA model using the fused dataset obtained the best performances with classification accuracy of 91.4 % in validation sets. Furthermore, spectral angle mapping (SAM) was performed to identify the infected tissue in wheat leaves 2 dai, and to monitor the pathogenesis of powdery mildew over time. The results from this study could be used to develop a portable field monitoring sensor for wheat powdery mildew.