小波特征与传统光谱特征估测冬小麦条锈病病情严重度的对比研究

为探讨通过小波特征监测小麦条锈病发病程度的可行性，利用连续小波变换提取的小麦冠层光谱350~1 300 nm范围内的9个小波特征和传统光谱特征（植被指数、一阶微分变换特征和连续统特征），借助偏最小二乘回归（PLSR）建立反演模型，分别将传统光谱特征（SFs）、小波特征（WFs）及传统光谱特征与小波特征结合(SFs & WFs）作为模型的输入，对小麦条锈病病情进行反演。结果表明：（1）小波特征与条锈病严重度的相关性比传统光谱特征强；（2）基于小波特征的模型估测精度（R为0.837）优于基于传统光谱特征的模型估测精度（R为0.824）；（3）传统光谱特征与小波特征结合的模型精度最高，R为0.876，RMSE仅为0.096，因而传统光谱特征与小波特征相结合能够更好地对小麦条锈病病情严重度进行估测。

Comparison of Wavelet Features and Conventional Spectral Features on Estimating Severity of Stripe Rust in Winter Wheat

The canopy reflectance of winter wheat infected by stripe rust under different severity conditions with artificial inoculation and the disease index (DI) were investigated in the field respectively.The partial least squares regression (PLSR) was applied to establish inversion model, through 12 conventional spectral features and 9 wavelet features in canopy spectra of 350~1 300 nm using continuous wavelet transform (CWT).The accuracy of wavelet features, conventional spectral features and combination of them was analyzed.It was indicated that the accuracy of wavelet features was superior to the traditional spectral features with R of 0.837; and the accuracy of the combination is the best with R of 0.876 and RMSE of 0.096.To a certain extent, the wavelet features could improve the estimation accuracy of the regression models.It has great potential for the application in monitoring wheat disease severity using hyperspectral remote sensing.