Understanding the changes in millable oil content and fatty acid composition of infrared radiation treated mustard (Brassica juncea) seeds using Vis-NIR-SWIR hyperspectral imaging

Abstract

Enhancing the recovery of edible oil during mechanical oil expression offers a potential solution to mitigate the gap between the supply and demand. Infrared (IR) pre-treatment has demonstrated an enhanced availability of mechanically extractable oil from mustard seeds. In this study, hyperspectral imaging (HSI) techniques were employed to investigate the spatial distribution of oil in mustard seeds subjected to IR treatment under varying voltage and time of exposure. Both visible near-infrared (Vis-NIR, 399–1003 nm) and short-wave infrared (SWIR, 895–1712 nm) HSI systems were utilized to acquire spectral data followed by chemometric analysis, including partial least square discriminant analysis (PLS-DA) and regression (PLSR), to develop prediction models. The PLS-DA model exhibited robust classification capabilities for the differentiating mustard seeds based on various IR treatments from control samples. Notably, the accuracy levels achieved were 94.7 % for Vis-NIR-HSI and 99.2 % for SWIR-HSI. The prediction of oil content and fatty acid components, such as erucic acid, oleic acid, saturated fatty acids and mono fatty acids (MUFAs) could be carried out by the PLSR model developed using SWIR-HSI spectral data (R2 > 0.80). These predictions closely aligned with the outcomes obtained from analytical techniques. However, when utilizing Vis-NIR spectral data, the predictions (R2 > 0.65) were comparatively less accurate.