Fusarium Head Blight (FHB) is an important fungal disease in corn mainly produced by F. graminearum and F. culmorum resulting in huge economic loss to the producers. Also, these Fusarium species further produces secondary metabolites called mycotoxins, predominantly deoxynivalenol (DON). DON is a stable mycotoxin found in corn grains, corn feed and processed corn products for human consumption. DON is a strong inhibitor of protein synthesis and affects translation and other ribosome functions in eukaryotic cells. DON is also shown to affect the immune system inducing vomiting, feed refusal and reduced growth in cattle and swine. Traditional analytical methods for DON quantification proved time consuming and expensive. This prompted the need for development of a non-destructive method to detect DON in corn grains and flour. In this study, near infrared (NIR) reflectance hyperspectral imaging in the spectral range between 900 and 2500nm was assessed for its potential to quantify DON in corn flour. Hyperspectral images of corn flour samples spiked with DON at five concentrations (0.5, 1, 2, 5 and 10 ppm) were acquired and the reference DON content in each sample was determined by ELISA method. Regression and classifications methods including linear regression, simple linear regression, sequential minimal optimization (SMO) and random forest were applied to fit the spectral data to the actual DON content in flour samples. The best regression was obtained for random forest classifier with R2 values of 0.93 for cross validation and 0.83 for prediction.

Pesticides are commonly used in yellow pea production and plays a positive role in reducing the crop yield losses caused by crop diseases, weeds, and pests. Glyphosate is the most widely used herbicides both as a desiccant and a weed control. The massive concentration of glyphosate found in yellow pea isolate is a major health problem for people worldwide. Currently the common practise for monitoring glyphosate residues in flour samples relies on chromatography and immunosorbent assays. These analysis uses a representative sample from each batch, requires long hours for estimation and is highly labour intensive. Therefore, in this study, the potential of near-infrared (NIR) hyperspectral imaging in the wavelength range between 900 and 2500 nm in predicting glyphosate content in yellow pea isolate was examined. The yellow pea isolate was spiked with glyphosate at five concentration levels (0 ppm, 5 ppm, 10 ppm, 15 ppm and 20 ppm) and used for this study. ELISA was used in the reference glyphosate investigation. Standard normal variate (SNV) preprocessing was used on the spectral data to minimize the impact of light scattering. To construct the partial least square regression (PLSR) model, the spectral data was correlated with the estimated reference glyphosate concentration of yellow pea isolate samples. The constructed model resulted in a coefficient of determination of calibration and prediction as 0.99 and 0.92 respectively. Hence, it was demonstrated that the NIR hyperspectral imaging system could be utilized in flour processing facilities to measure the amounts of glyphosate residue in yellow pea isolate.