Adulteration Level Prediction and Screening in Bulk Rice Using Laser-Induced Breakdown Spectroscopy Coupled with Nonlinear Modeling Academic Article in Scopus uri icon

abstract

  • Rice authenticity verification is crucial for ensuring food quality and maintaining consumer trust. In this study, laser-induced breakdown spectroscopy (LIBS) combined with advanced nonlinear modeling was explored for the semiquantitative prediction and classification-based screening of adulteration levels in bulk rice. A total of 240 rice samples, comprising authentic grains and those adulterated at varying levels, were analyzed using LIBS. After careful spectral preprocessing to reduce noise and multicollinearity, correlation analysis and feature selection identified four key emission lines: C I (28.8% importance), Mg II (26.0%), Fe I (25.3%), and Na I (19.9%). These elemental markers served as predictors to develop decision tree (DT), extreme gradient boosting (XGB), and support vector regression (SVR) models aimed at assessing rice authenticity. Although predictive performance was lower in the testing set (R2 = 0.55) compared to the training set (R2 = 0.91), XGB achieved the most reliable predictions by capturing subtle relationships between elemental composition and adulteration levels. Classification into predefined adulteration categories yielded even better outcomes, reaching balanced accuracies of 90% for the training samples and 85% for the testing samples, with performance improving as the degree of adulteration increased. Complementary principal component analysis (PCA) further confirmed clear differentiation among authentic and adulterated rice, even in morphologically similar varieties. The LIBS-XGB framework thus offers a rapid, minimally invasive, and cost-effective strategy for routine rice authenticity screening, supporting enhanced traceability and quality control within cereal supply chains. © 2025 American Chemical Society

publication date

  • November 21, 2025