Discrimination between durum and common wheat kernels using near infrared hyperspectral imaging

• Published in: Journal of cereal science Vol. 84; pp. 74 - 82
• Main Authors: Vermeulen, Philippe, Suman, Michele, Fernández Pierna, Juan Antonio, Baeten, Vincent
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.11.2018
• Subjects: Durum wheat; Fraud detection; Morphology; Near infrared hyperspectral imaging; Durum wheat; Near infrared hyperspectral imaging; Fraud detection; Morphology
• ISSN: 0733-5210; 1095-9963
• DOI: 10.1016/j.jcs.2018.10.001

According to Italian regulation, 3% of common wheat - CW (Triticum aestivum) in durum wheat - DW (Triticum durum) is the maximum permitted to produce pasta. Therefore, efficient methods for the detection of accidental or intentional contamination of DW products with CW are required. Until now, all the studies dealing with the detection of CW in DW have been mainly based on macroscopic, microscopic or molecular biology methods. In this recent work, near infrared (NIR) hyperspectral imaging was evaluated as a tool for discriminating between both species of wheat at the singulated kernel and bulk sample levels. This study involved the analysis of 77 samples of DW and 180 samples of CW. NIR images were acquired on a total of 4112 kernels at kernel level and on a total of approximately 51.4 kg of kernels at bulk level. To discriminate DW from CW, four approaches were studied based on morphological criteria, NIR spectral profile, protein content criteria and ratio of vitreous/non-vitreous kernels. Partial least squares discriminant analysis was used as a classification method for the construction of the discrimination models. Results showed that a combination of morphological and NIR spectral approaches could detect fraud in sample classification with 99% accuracy. •Solutions for sorting kernels at the entry point of the production chain.•Near infrared hyperspectral imaging for discriminating between wheat species.•The criteria are morphology, near infrared spectra, protein and vitreousness.•Data fusion by combining the four criteria provides new discrimination indicators.