Detection of rice leaf folder, Cnaphalocrocis medinalis (Guenée) (Lepidoptera: Crambidae) infestation using ground-based hyperspectral radiometry

Abstract

Hyperspectral remote sensing is a useful technique for detecting spatio-temporal changes in crop morphological and physiological health. In order to identify the pest-sensitive bands for rice leaf folder (RLF), the ground-based hyperspectral data were recorded at varying damage levels. The first and second-order derivatives were correlated with correlation coefficient r and per cent leaf damage. The common region identified were recognized as sensitive regions (508–529, 671–680, 721–759, 779–786 and 804–820 nm). The absorption dips were also found using Sensitivity and Continuum Removal Analysis. Combining all, a total of nine spectral bands (518, 549, 661, 674, 678, 731, 789, 816 and 898 nm) were identified as pest-sensitive bands for RLF. The feature-selection method was employed using RELIEFF algorithm to find out the band combinations and bands 518, 661 and 731 nm yielded maximum accuracy of 81.67%