Impact of spectral interval on wavelet features for detecting wheat yellow rust with hyperspectral data

Detection of yellow rust using hyperspectral data is of practical importance for disease control and prevention. As an emerging spectral analysis method, continuous wavelet analysis (CWA) has shown great potential for the detection of plant diseases and insects. Given the spectral interval of airborne or spaceborne hyperspectral sensor data differ greatly, it is important to understand the impact of spectral interval on the performance of CWA in detecting yellow rust in winter wheat. A field experiment was conducted which obtained spectral measurements of both healthy and disease-infected plants. The impacts of the mother wavelet type and spectral interval on disease detection were analyzed. The results showed that spectral features derived from all four mother wavelet types exhibited sufficient sensitivity to the occurrence of yellow rust. The Mexh wavelet slightly outperformed the others in estimating disease severity. Although the detecting accuracy generally declined with decreasing of spectral interval, relatively high accuracy levels were maintained (R2>0.7) until a spectral interval of 16 nm. Therefore, it is recommended that the spectral interval of hyperspectral data should be no larger than 16 nm for the detection of yellow rust. The relatively loose spectral interval requirement permits extensive applications for disease detection with hyperspectral imagery.