Non-destructive assessment of chilling injury in red pepper powder using short-wave-infrared and XGBoost algorithm

This study aimed to evaluate red pepper powder quality by the extent of chilling injury and develop a method for detecting chilling injury-affected pepper powder. Pepper powder produced from chilling injury-affected pepper fruits exhibited increased bitter amino acids, microbial counts, and biogenic amines and decreased sweetness index and organic acid levels. These quality deteriorations indicate the need to detect chilling injury in pepper powders. The color values were insufficient to identify the occurrence of chilling injury. Therefore, hyperspectral imaging was used to detect the chilling injury. Based on the feature importance metrics results from XGBoost and correlation analysis, five key wavelengths were selected from a total of 188 wavelengths. The XGBoost model, using selected wavelengths, demonstrated higher accuracy (100 %) in discriminating the chilling injury level in pepper powder compared to that using full-wavelengths (98.0 %). These results provide insights into the rapid and non-destructive quality assessment of pepper powder.