The hyperspectral bands sensitive to the disease severity levels of wheat powdery mildew was elucidated in this study. The disease severity levels of wheat powdery mildew were also inverted by the extracting characteristic parameters, which provided a basis for detecting the wheat powdery mildew using hyperspectral data. The spectral data of single leaves was obtained at heading stage, the spectral characteristic parameters and sensitivity of wheat leaves were analyzed qualitatively and quantitatively. The result showed that spectral reflectivity within the visible wavebands (400—760 nm) was increased with the aggravating disease severity levels. The spectral sensitivity reached the maximum value within visible wavebands and the optimal sensitive bands for detecting disease severity levels was 630—680nm. After the spectrum was continuum removal-treated, the absorption position moved to longer wavelength with the aggravating disease severity levels and the disease severity levels showed extremely significant negative correlations with the absorption height, absorption width and absorption area. The linear regression equation has high determination coefficient and low root mean square error using the right AAI as independent variable to establish the model. Moreover, the precision verification test also showed that the model performed best in monitoring wheat powdery mildew. In conclusion, disease severity levels of wheat powdery mildew could be inverted effectively by hyperspectral technology, which provides the foundation for detecting wheat powdery mildew.