Rice sheath blight severity monitoring method based on XGBoost

The invention provides a rice sheath blight severity monitoring method based on XGBoost. The method comprises the following steps of: representing disease information by coupling remote sensing and meteorological indexes; performing dimension reduction on remote sensing and meteorological characteristics by using PCA; and finally, constructing a monitoring model by using an XGBoost method, training a next weak learner by the XGBoost through optimizing the gradient of a loss function in each round of iteration, and adding the next weak learner into the model to reduce the residual error of the previous round. A rice sheath blight disease monitoring model is constructed through XGBoost algorithm optimization, parameter optimization is carried out on the model through three methods of GS, GA and PSO, after optimization, the AUC value is 0.792, and the TSS value is 0.621. Compared with a monitoring model constructed by default parameters, the AUC value is improved by 0.077, and the TSS value is improved by 0.138.