1Department of Biosystems Engineering, University of Manitoba, Winnipeg, MB, Canada, R3T 5V6
2Agriculture and Agri-Food Canada, c/o Department of Biosystems Engineering, University of Manitoba, Winnipeg, MB, Canada, R3T 5V6
3Canadian Grain Commission, c/o Richardson Centre for Functional Foods and Nutraceuticals, University of Manitoba, Winnipeg, MB, Canada, R3T 2E1
*Email: Digvir.Jayas@umanitoba.ca (corresponding author)
Online published on 20 May, 2016.
Near-Infrared (NIR) hyperspectral imaging system operated in the wavelength region of 1000 to 1600 nm was used to detect ochratoxin A contaminated barley kernels. Five concentration levels of ochratoxin A contaminated barley kernels, five infection periods of non-ochratoxin A producing Penicillium verrucosum infected barley kernels, and sterile control samples were subjected to single kernel imaging. The acquired image data were processed and subjected to Principal Component Analysis (PCA) to identify significant wavelengths. Linear and quadratic discriminant pair-wise, two-class, and six-class models were developed and tested with statistical and histogram features extracted from significant wavelengths. The significant wavelength 1480 nm was identified only in the ochratoxin A contaminated samples, and not identified in the non-ochratoxin A producing P. verrucosum infected samples. The significant wavelengths 1310, and 1360 nm were found in both ochratoxin A contaminated samples and non- ochratoxin A producing P. verrucosum infected ones. The classification accuracy of more than 97% was achieved in discriminating between different concentration levels of ochratoxin A contaminated samples and a classification accuracy of 100% was achieved in discriminating between sterile control samples, and ochratoxin A contaminated samples.
Penicillium verrucosum, Near-Infra Red (NIR), wavelengths, significance, Principal Component Analysis (PCA), Ochratoxin A, detection