Remote sensing index selection of leaf blight disease in spring maize based on hyperspectral data

Abstract: Leaf blight is one of the major diseases of spring corns. Analyzing crop canopy spectral features and establishing remote sensing monitoring indices by employing the method of ground spectrum test are the foundation for implementing regional disease remote sensing monitoring, and the major basis for designing satellite sensor spectrum. By taking 4 varieties of spring corns i.e. highly resistant, resistant, infected, and highly infected spring corns as the study objects, this paper designed an artificially controlled plot experiment in Meixian County, Shannxi Province. Through artificial inoculation of leaf blight spores with different concentrations, the study established 4 disease infected land plots including normal corn, mildly infected corn, moderately infected corn, and severely infected corn, and conducted ground hyperspectral observation of 4 development stages i.e. tasseling stage, silking stage, milk-ripe stage and ripe stage of spring corns. In order to realize the remote sensing monitoring on leaf blight of spring corns, based on the spring corn canopy spectral data, this paper analyzed the change features of spring corn canopy spectral reflectance and first derivative spectral value of crop areas with different disease severities and at different growth stages, and identified the sensitive band range of spring corn leaf blight and proper disease monitoring period; meanwhile the study established a special remote sensing index for spring corns based on the spectral features of sensitive wave bands, namely, the first derivative of spectral in red edge core area. Finally, to validate the effectiveness of the index proposed in this paper, the study used a total of 180 spectral observation samples in the 4 crop areas, which were obtained by using the crossing sampling method, and made a comparison between the indices and the other commonly used disease monitoring indices in terms of their correlation with disease severity. The result showed that, spring corn leaf blight with different order of severity could be more significantly represented during the milk-ripe stage of the spring corn. Along with the increase of disease severity, the reflectance in near-infrared band decreased gradually, and showed a change of gradient, which was suitable for the leaf blight remote sensing monitoring and classification of disease severity; response of first derivative of spring corn canopy spectrum was relatively sensitive, especially within the range of red edge core area (725-740 nm). There was a significant monotonous change relation between first derivative of spectrum and disease severity, showing a very significant negative correlation. The experimental result also showed that there was a relatively high correlation between the remote sensing monitoring index proposed in this paper and the disease index, with the correlation coefficient of 0.995 0. Classification accuracy of different disease severity reached 100%, and the dispersion degree of index values was lower than that of other commonly used monitoring indices, with a higher distribution stability, indicating that the indices proposed in this paper can be applied in remote sensing monitoring operations.