By combining the spectral and texture features of images captured by unmanned aerial vehicles (UAVs), the accurate and timely detection of wheat Fusarium head blight (FHB) can be realized. This study presents a methodology to select the optimal window size of the gray-level co-occurrence matrix (GLCM) to extract texture features from UAV images for FHB detection. Host conditions and the disease distribution were combined to construct the model, and its overall accuracy, sensitivity, and generalization ability were evaluated. First, the sensitive spectral features and bands of the UAV-derived hyperspectral images were obtained, and then texture features were selected. Subsequently, spectral features and texture features extracted from windows of different sizes were input to classify the area of severe FHB. According to the model comparison, the optimal window size was obtained. With the collinearity between features eliminated, the best performance of the logistic model reached, with an accuracy, F1 score, and area under the receiver operating characteristic curve of 0.90, 0.79, and 0.79, respectively, when the window size of the GLCM was 5×5 pixels on May 3, and of 0.90, 0.83, and 0.82, respectively, when the size was 17×17 pixels on May 8. The results showed that the selection of an appropriate GLCM window size for texture feature extraction enabled more accurate disease detection.

Remote sensing technology provides a feasible option for early prediction for wheat Fusarium head blight (FHB). This study presents a methodology for the dynamic prediction of this classic meteorological crop disease. Host and habitat conditions were comprehensively considered as inputs of the FHB prediction model, and the advantages, accuracy, and generalization ability of the model were evaluated. Firstly, multi-source satellite images were used to predict growth stages and to obtain remote sensing features, then weather features around the predicted stages were extracted. Then, with changes in the inputting features, the severity of FHB was dynamically predicted on February 18, March 6, April 23, and May 9, 2017. Compared to the results obtained by the Logistic model, the prediction with the Relevance Vector Machine performed better, with the overall accuracy on these four dates as 0.71, 0.78, 0.85, and 0.93, and with the area under the receiver operating characteristic curve as 0.66, 0.67, 0.72, and 0.75. Additionally, compared with the prediction with only one factor, the integration of multiple factors was more accurate. The results showed that when the date of the remote sensing features was closer to the heading or flowering stage, the prediction was more accurate, especially in severe areas. Though the habitat conditions were suitable for FHB, the infection can be inhibited when the host’s growth meets certain requirements.