Scholarlink

<p>This study proposes a new method to identify adulteration of transparent liquid food by optical method. Cane sugar adulteration (0–10% w/w) was performed in commercial 100% clear apple juice. Samples were illuminated by six laser beams (532, 635, 780, 808, 850, and 1064&#xa0;nm) using laser-induced diffuse reflectance imaging (LIDRI) and laser-induced diffuse transflectance imaging (LIDTI). A systematic comparison between the two configurations was conducted to evaluate their relative performance. Backscattering features were extracted using unimodal (for LIDRI) and bimodal (for LIDTI) Gaussian curve-fitting. The Kruskal–Wallis test showed that the extracted features significantly responded to sugar addition (<i>p</i> &lt; 0.05). The LIDTI features exhibited greater sensitivity to adulteration than those of LIDRI. Random forest (RF), k-nearest neighbors (kNN), and support vector machine (SVM) algorithms were applied for classification and regression. LIDTI outperformed LIDRI in both tasks. The best performance was achieved using RF for classification (cross-validation accuracy of 98.9% and prediction accuracy of 100%) and kNN for regression (<i>R</i><sup>2</sup><sub>CV</sub> = 0.933 and <i>R</i><sup>2</sup><sub>P</sub> = 0.937). These results demonstrate that LIDTI is a promising approach for rapid and non-destructive detection of sugar adulteration in clear apple juice.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Comparison of Diffuse Reflectance and Transflectance Imaging for Detecting Cane Sugar Adulteration in Clear Apple Juice

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月30日

查看全文
  • Hoa Xuan Mac,
  • Nga Thi Thanh Ha,
  • László Ferenc Friedrich,
  • Lien Le Phuong Nguyen,
  • László Baranyai

摘要

This study proposes a new method to identify adulteration of transparent liquid food by optical method. Cane sugar adulteration (0–10% w/w) was performed in commercial 100% clear apple juice. Samples were illuminated by six laser beams (532, 635, 780, 808, 850, and 1064 nm) using laser-induced diffuse reflectance imaging (LIDRI) and laser-induced diffuse transflectance imaging (LIDTI). A systematic comparison between the two configurations was conducted to evaluate their relative performance. Backscattering features were extracted using unimodal (for LIDRI) and bimodal (for LIDTI) Gaussian curve-fitting. The Kruskal–Wallis test showed that the extracted features significantly responded to sugar addition (p < 0.05). The LIDTI features exhibited greater sensitivity to adulteration than those of LIDRI. Random forest (RF), k-nearest neighbors (kNN), and support vector machine (SVM) algorithms were applied for classification and regression. LIDTI outperformed LIDRI in both tasks. The best performance was achieved using RF for classification (cross-validation accuracy of 98.9% and prediction accuracy of 100%) and kNN for regression (R2CV = 0.933 and R2P = 0.937). These results demonstrate that LIDTI is a promising approach for rapid and non-destructive detection of sugar adulteration in clear apple juice.