Multi-temporal wheat disease detection by multi-spectral remote sensing

摘要: For the implementation of site-specific fungicide applications, spatio-temporal dynamics crop diseases must be well known. Remote sensing can a useful tool to monitor heterogeneity vitality within agricultural sites. However, identification fungal infections at an early growth stage is essential. This study examines potential multi-spectral remote for multi-temporal analysis diseases. Within experimental field, 6 ha plot winter wheat was grown, containing all possible infective stages powdery mildew (Blumeria graminis) and leaf rust (Puccinia recondita) pathogens. Three high-resolution images were used execute infection dynamics. A decision tree, using mixture tuned matched filtering (MTMF) results Normalized Difference Vegetation Index (NDVI), applied classify data into areas showing different levels disease severity. Classification compared ground truth data. The classification accuracy first scene only 56.8%, whereas scenes from May 28th June 20th achieved considerably higher accuracies 65.9% 88.6% respectively. showed that are generally suitable detect in-field heterogeneities vigour but moderately detection infections.