Utilising near-infrared hyperspectral imaging to detect low-level peanut powder contamination of whole wheat flour. (August 2019)

• Record Type: Journal Article
• Authors: Zhao, Xin; Wang, Wei; Ni, Xinzhi; Chu, Xuan; Li, Yu-Feng; Lu, Chengjun
• Is Part Of: Biosystems engineering. Volume 184(2019)
• Page Start: 55
• Page End: 68
• Publication Date: 2019-08
• Subjects: Near-infrared hyperspectral imaging -- Low-level contamination -- Peanut powder -- Whole wheat flour -- Visualisation; Bioengineering -- Periodicals; Agricultural engineering -- Periodicals; Biological systems -- Periodicals; Génie rural -- Périodiques; Systèmes biologiques -- Périodiques; 631
• DOI: 10.1016/j.biosystemseng.2019.06.010
• Languages: English
• ISSNs: 1537-5110

Abstract

Abstract : Near-infrared hyperspectral imaging (HSI) was used for detecting low levels of peanut powder contamination in whole wheat flour, with concentrations of 0.01–10% (w/w). Two types of whole wheat flours, i.e. spring wheat flour (WFS) and winter wheat flour (WFW), were used. Minimum noise fraction combined with n-Dimensional visualiser tool was applied on light intensity calibrated hyperspectral images for preliminary discrimination. Competitive adaptive reweighted sampling (CARS) was applied for optimal wavelength selection. Partial least squares regression (PLSR) models with standard normal variate followed by Savitzky–Golay first derivatives had the best performance, with coefficients of determination of prediction ( R 2 p ) of 0.993 and 0.991, and root mean square error of prediction (RMSEP) of 0.251% and 0.285%, respectively for contaminated WFS and WFW samples. Prediction maps based on PLSR models permitted visualising spatial variations in the concentration of peanut contamination. The results indicated that near-infrared HSI has the potential to detect low-level peanut contamination in whole wheat flour. Highlights: A light intensity calibration method for hyperspectral image de-noising was used. Spectral pretreatment by SNV + SGD and CARS wavelength selection gave best PLSR model. Peanut powder concentrations as low as 0.3% were detected in spring wheat flour. Peanut powder concentrations as low as 0.5% were detected in winter wheat flour. Prediction map permitted visualising the spatial variation of concentrations ≥3%.