Detection of aphid infestation on pakchoi (Brassica chinensis) based on visible/near-infrared hyperspectral imaging technology

Abstract. Green peach aphid is a major pest on pakchoi which could cause serious decline in quality and yield. This study employed hyperspectral imaging system to detect incipient infestation of aphid and identify different infestation stages. Chlorophyll fluorescence technology was employed to estimate the actual damage caused by aphid. Visible/near-infrared hyperspectral images (380 to 1030 nm) of pakchoi leaves infested by aphid were acquired and divided into five stages according to infestation days. SVM was employed as classifier, and MCUVE, CARS, VCPA, MCFS and SPA were compared to select the best optimal characteristic wavelengths. The detection accuracy based on full band of the last three stages was 100% and accuracy of the first two stages was relatively low. MCFS acquired the best overall accuracy of 97.94% and 95.54% for train set and test set respectively among five selection methods. Chlorophyll fluorescence parameters of QY_max and qL_D2 confirmed that the damage to photosynthetic system in leaf was not serious in the first two stages. Thus, overall results indicated that hyperspectral imaging system could be used to detect incipient aphid infestation effectively.