Scholarlink

<p>This study presents a hyperspectral imaging (HSI) system integrated with classification techniques to assess the authenticity and detect adulteration in Iranian rice samples. The focus was on Hashemi rice, a premium and costly variety, which was intentionally mixed with Neda and Shiroudi rice varieties that share similar morphological features but are of lower quality and price. Additionally, Fajr rice, a domestic variety, was adulterated with an imported variety exhibiting comparable shape characteristics but inferior quality and cost. Initially, Principal Component Analysis (PCA) was applied to the preprocessed hyperspectral data, effectively enabling visual differentiation among all sample groups. Evolutionary Wavelength Selection (EWS) was used for selecting the more effective spectral bands. For classification, Partial Least Squares-Discriminant Analysis (PLS-DA) and deep learning-based Convolutional Neural Networks (CNN) were employed. Results demonstrated that the accuracy values of both models exceeded 90%. The CNN model slightly outperformed the PLS-DA approach. It was ultimately utilized to detect adulteration in Hashemi and Fajr samples, successfully identifying levels of adulteration as low as 5%. R² values were higher than 0.85 and 0.99 when Neda and Shiroudi varieties were added to the Hashemi variety, respectively, and 0.92 when the imported variety was added to the Fajr variety. Overall, the HSI + CNN algorithm provided a rapid and non-invasive solution for evaluating rice authenticity.</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

Rice authenticity evaluation and adulteration detection using hyperspectral imaging system coupled with deep learning

  • 期刊论文

  • 发表日期：

    2025年12月2日

查看全文
  • Mahsa Edris,
  • Sajad Kiani,
  • Mahdi Ghasemi-Varnamkhasti,
  • Hassan Yazdanpanah,
  • Zahra Izadi

摘要

This study presents a hyperspectral imaging (HSI) system integrated with classification techniques to assess the authenticity and detect adulteration in Iranian rice samples. The focus was on Hashemi rice, a premium and costly variety, which was intentionally mixed with Neda and Shiroudi rice varieties that share similar morphological features but are of lower quality and price. Additionally, Fajr rice, a domestic variety, was adulterated with an imported variety exhibiting comparable shape characteristics but inferior quality and cost. Initially, Principal Component Analysis (PCA) was applied to the preprocessed hyperspectral data, effectively enabling visual differentiation among all sample groups. Evolutionary Wavelength Selection (EWS) was used for selecting the more effective spectral bands. For classification, Partial Least Squares-Discriminant Analysis (PLS-DA) and deep learning-based Convolutional Neural Networks (CNN) were employed. Results demonstrated that the accuracy values of both models exceeded 90%. The CNN model slightly outperformed the PLS-DA approach. It was ultimately utilized to detect adulteration in Hashemi and Fajr samples, successfully identifying levels of adulteration as low as 5%. R² values were higher than 0.85 and 0.99 when Neda and Shiroudi varieties were added to the Hashemi variety, respectively, and 0.92 when the imported variety was added to the Fajr variety. Overall, the HSI + CNN algorithm provided a rapid and non-invasive solution for evaluating rice authenticity.