Scholarlink

<p>The development of non-destructive sensing technologies has significantly improved quality evaluation in the agri-food sector by enabling rapid and contact-free assessment of both internal and external attributes of fruits and vegetables. This review compares three widely used modalities—hyperspectral imaging (HSI), thermal imaging (TI), and visible–near infrared spectroscopy (Vis–NIRS)—with attention to their working principles, system configurations, spectral properties, and predictive performance. Their effectiveness in estimating important physicochemical parameters, including soluble solids content, firmness, moisture, and pigment composition, is examined using commonly reported evaluation metrics such as the coefficient of determination (R<sup>2</sup>), root mean square error (RMSE), and residual predictive deviation (RPD). In addition, the contribution of chemometric and data-driven modeling approaches to improving prediction accuracy and robustness is discussed. Vis–NIRS is well suited for rapid compositional analysis, HSI provides detailed spatial and spectral information, and TI offers complementary insights into surface temperature variations associated with defects or physiological changes. The integration of these techniques through multimodal frameworks presents promising opportunities for practical grading and quality monitoring applications. Future developments are expected to focus on multimodal data fusion frameworks, deep learning–based feature extraction, transfer learning–enabled model generalization, and the deployment of portable, real-time sensing systems for in-field applications.</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

Comparative Evaluation of Hyperspectral, Thermal, and Near-Infrared Imaging Systems for Non-Destructive Prediction of Fruit and Vegetable Quality: A Review

  • 期刊论文

  • 发表日期：

    2026年5月18日

查看全文
  • Kok Jeng Lim,
  • Phebe Ding,
  • Nazmi Mat Nawi,
  • Mashitah Jusoh,
  • Zailani Khuzaimah

摘要

The development of non-destructive sensing technologies has significantly improved quality evaluation in the agri-food sector by enabling rapid and contact-free assessment of both internal and external attributes of fruits and vegetables. This review compares three widely used modalities—hyperspectral imaging (HSI), thermal imaging (TI), and visible–near infrared spectroscopy (Vis–NIRS)—with attention to their working principles, system configurations, spectral properties, and predictive performance. Their effectiveness in estimating important physicochemical parameters, including soluble solids content, firmness, moisture, and pigment composition, is examined using commonly reported evaluation metrics such as the coefficient of determination (R2), root mean square error (RMSE), and residual predictive deviation (RPD). In addition, the contribution of chemometric and data-driven modeling approaches to improving prediction accuracy and robustness is discussed. Vis–NIRS is well suited for rapid compositional analysis, HSI provides detailed spatial and spectral information, and TI offers complementary insights into surface temperature variations associated with defects or physiological changes. The integration of these techniques through multimodal frameworks presents promising opportunities for practical grading and quality monitoring applications. Future developments are expected to focus on multimodal data fusion frameworks, deep learning–based feature extraction, transfer learning–enabled model generalization, and the deployment of portable, real-time sensing systems for in-field applications.