A NOVEL METHOD FOR DETECTION OF PIERIS RAPAE LARVAE ON CABBAGE LEAVES USING NIR HYPERSPECTRAL IMAGING

摘要

Pest detection is important in agricultural production. In this study, near-infrared (NIR) hyperspectral imaging technology was applied to detect 4th instar Pieris rapae larvae on cabbage leaves. After hyperspectral imaging acquisition (1000-1600 nm), successive projections algorithm (SPA) was implemented to select effective wavelengths (EWs). Partial least squares discriminant analysis (PLS-DA) and back-propagation neural network (BPNN) models were developed based on 13 selected EWs to distinguish between leaves and larvae, both yielding acceptable results of correlation coefficient in the calibration set (RC) above 0.98 and classification accuracy in the prediction set above 96%. In terms of computation time, the developed SPA-PLS-DA model was chosen for pixel-wise detection of larvae in mixed samples and achieved accurate visual results. The promising results indicated that it is feasible to use NIR hyperspectral imaging to detect Pieris rapae larvae accurately and intuitively. This technology will be helpful in the early pest control.