On-line detection of toxigenic fungal infection in wheat by visible/near infrared spectroscopy

Efficiently identifying and separating grains infected by toxigenic fungi can prevent mycotoxins entering in food chain. This study presents an on-line scheme for synchronous recognition of hazardous fungal contamination as well as prediction of infection level in wheat based on visible/near infrared (Vis/NIR) spectroscopy. Sterilized wheat kernels contaminated with Fusarium and Aspergillus strains were scanned in wavelength range from 600 to 1600 nm under on-line condition. Spectral analysis and principal component analysis (PCA) indicated fungi activity could be monitored by Vis/NIR spectroscopy. Subsequently, linear discriminant analysis (LDA) modeling obtained correct classified rate of 91.7% for classification of samples infected by different fungal strains after stored for 3 d and 88.3% for discrimination of samples with different infection levels (acceptable, slightly moldy and highly moldy). Moreover, partial least squares regression (PLSR) modeling achieved good quantification results for prediction of colony counts in samples (Rp2 = 0.890, RMSEP = 0.369 log CFU/g, RPD = 3.03). These findings verify the potential of Vis/NIR spectroscopy for early detection of fungal contamination in grains during on-line processing.