Insect species and infestation level determination in stored wheat using near-infrared spectroscopy

摘要

Near-infiared spectroscopy (NIRS) technique was evaluated foi detection of four different life stages (i.e. eggs, larvae, pupae, and adults) of Sitophilus oryzae (rice weevil) in artificially infested bulk samples of Canada Western Red Spiing (CWRS) wheat at infestation levels from 0 through 50% at 5% infestation intervals. Sound wheat kernels were also infested with Rhyzopertha dominica (lesser grain borer) to create infestation levels of 0, 5,10,20,25,50, 75, and 100% and the spectral data from their pupae were compared to the pupal stage of rice weevil at corresponding levels of infestation. To distinguish wheat kernels infested with pupae of lesser grain borer from those infested with rice weevil, Principal Component Analysis (PCA) was employedand insect differentiation was done by examining score plots of spectral data. Original and difference spectra of infested wheat kernels and sound wheat kernels were tested. It was observed that differentiation of insect species becomes easier as infestation levels increase for difference spectra. Calibration models for quantitative determination of infestation levels were developed using Partial Least Square Regression (PLSR). Multiplicative Scatter Correction (MSC) and Standard Normal Variant (SNV) followed by detrending were found equally effective in removing irrelevant spectral information. Prediction performed well for high infestation levels but lower classification accuracies were obtained at low infestation levels. kw:near-infrated spectioscopy; wheat; Sitophilus oryzae; Rhyzopertha dominica; infestation; Principal Component Analysis; Partial Least Squares Regression; Multiplicative Scatter Correction; Standard Normal Variate and detrending; spectroscopie à infrarouge rapproché; blé; Sitophilus oryzae; Rhyzopertha dominica; infestation; analyse des composantes principales; régression des moindres canes partiels; coirélation des écarts multiplicatifs; variante normale standard et désoiganisation