小麦条锈病和白粉病作为我国麦区两种重要病害, 在田间常同时发生, 为病害防治管理带来困难。 基于实验测试获得白粉病、 条锈病叶片光谱数据, 探讨采用光谱分析对两种病害进行区分识别及严重度监测的可行性。 通过相关分析和独立T检验, 筛选出对白粉病和条锈病敏感度差异较显著的波段及光谱特征, 包括665～684, 718～726 nm等6个波段范围, 以及DEP550-770, SIWSI等11个光谱特征。 基于这些波段和特征, 采用FLDA构建病害判别模型; 借助PLSR分析构建病情严重度反演模型。 研究结果表明, 筛选得到的反射率波段和光谱特征能够较好地区分两种病害, 判别模型总体精度达到80%以上, 准确度较高。 其中, 染病比率超过20%的病叶区分和识别精度可达95%。 同时, 分别基于两种病害敏感光谱特征构建的病情严重度反演模型能够较好地估测病情严重度, 两种病害估测均方根误差均低于15%。 上述叶片尺度小麦白粉病和条锈病区分和严重度反演模型为进一步研究两种病害冠层尺度的区分和监测提供基础。

Yellow rust and powdery mildew are two important diseases of winter wheat in China. The coincidence of their occurrence in field poses a challenge in their management and prevention. In the present study, the leaf spectra of the two diseases were measured by a spectrometer. Based on these data, we assessed the feasibility of differentiating the two diseases and evaluating their severity degrees. The disease sensitive bands and spectral features were identified through correlation analysis and independent t-test, including band regions at 665～684, 718～726 nm etc. and spectral features of DEP550-770, SIWSI etc. Based on these bands and spectral features, the models for disease discrimination and severity retrieving were developed according to FLDA and PLSR analysis, respectively. The results showed that the selected bands and spectral features can differentiate the yellow rust and powdery mildew explicitly, which yielded an OAA of 80%. It is noted that the discrimination model performed especially well (OAA=95%) in classifying those diseased leaves with damage proportion over 20%. The retrieving model of disease severity that were constructed by spectral features achieved reasonable estimates, with the RMSE for both diseases less than 15%. The leaf level models for discriminating powdery mildew and yellow rust and estimating their disease severity serve as a basis for further study in diseases differentiation and detection at canopy level.