基于叶片高光谱特征的小麦白粉病严重度估算模式

为了解白粉病胁迫下小麦叶片特征并预测其危害程度,基于大田小区和温室盆栽小麦白粉病接种试验,采用高光谱仪测定受白粉病不同程度危害的冬小麦叶片光谱反射率,并分析光谱特征参数与白粉病严重度间的关系。结果表明,随着小麦白粉病病情的加重,在可见光350~700nm波段内,叶片光谱反射率增加;而在700~1050nm近红外波段内,叶片光谱反射率明显降低。400~500nm和610~690nm为光谱敏感波段,在650~680nm波段相关系数最高(r0.75)。光谱参数MCARI、PSRI、VARIgreen和AI对叶片病害严重度拟合效果较好,决定系数(R2)变化范围为0.77~0.82,标准误差为9.34~10.14。模型检验表明,小麦单叶片病害严重度超过10%时,检验结果较为理想,单叶片病害严重度低于10%时,则定量估算误差偏大,10%严重度可作为光谱法识别小麦白粉病的临界值。光谱参数MCARI和VARIgreen对小麦白粉病反应敏感,估算误差较小,可作为小麦白粉病严重度的最佳估算模型。

Estimation Model of Wheat Powdery Mildew Severity Based on Leaves Hyperspectral Characteristics

Wheat Powdery mildew is a popular wider, onset heavier crop diseases all over the world, in order to pinpoint the leaf characteristics and endanger degree under the stress of wheat powdery mildew, after wheat powdery mildew inoculated in field plot and greenhouse potting, the spectrum reflectance of single leaves was measured and the severity level of powdery mildew was investigated in wheat field, and the relationships between spectral parameters and disease severity of powdery mildew were investigated. The results showed that spectrum reflectance increased significantly in visible light region 350~700 nm and obviously decreased in near infrared flat region 700~1 050 nm with the increase of the severity level of powdery mildew. The sensitivity bands occurred in visible light region of 400~500 nm and 610~690 nm mostly, and correlation coefficient over 0.75 was observed between the spectrum reflectance of 650~680 nm and the severity level of powdery mildew. Regression models with spectral variables of MCARI, PSRI, VARIgreen and AI produced better estimation of disease severity, with determination coefficient (R) of 0.77~0.82 and standard errors of 9.34~10.14. Testing with independent datasets indicated that the predicting models gave accurate estimation while the disease severity over 10% as ratio of leaf area infected to total leaf area, and gave poor estimation while the disease severity lower than 10%, so the disease severity of 10% could be used as a threshold for using the leaf spectrum reflectance as a diagnose index for wheat leaf infected by powdery mildew in early period. The models of MCARI and VARIgreen had best estimated precision, and were commended as the best models to estimate disease severity of powdery mildew in wheat.