Hyperspectral imaging techniques for detection of foreign materials from fresh-cut vegetables

Foreign materials (FMs) in fresh-cut vegetables are a huge concern for the fresh-cut industry since they affect product safety and quality. Therefore, effective methods of detecting FMs in industrial processing operations are urgently required. In this study, three hyperspectral imaging (HSI) techniques (VNIR, SWIR, and fluorescence) were investigated to distinguish the FMs from seven common fresh-cut vegetables. In addition, a partial least squares discriminant analysis (PLS-DA) model was developed for all types of vegetables to identify the FMs. Among the three different HSI systems, SWIR provided the best FMs detection accuracy (99 %), followed by VNIR (89 %) and fluorescence (64 %). Furthermore, the beta coefficient obtained from the PLS-DA model was applied to the hyperspectral image to reveal the FMs visually. Because the SWIR HSI system provided the best result, three variable selection techniques (SFS, SPA, and iPLS) were applied to select important wavelengths from the SWIR HSI data, and new PLS-DA models were developed. The results suggested that the SWIR HSI techniques with the SPA-PLS-DA model (99 % overall detection accuracy) could be efficiently utilized in in-dustrial applications for rapid and nondestructive FMs detection in fresh-cut vegetables.